Control Inhibitory in Young Children and Its Role in Emerging Internalization
Grazyna Kochanska, KatUeen Murray, Tanya Y. Jaccfnes, Amy L. Koenigr, and Kimberly A.
University of Iowa
KocHANSKA, GRAZYNA; MURRAY, KATHLEEN; JACQUES, TANYA Y.; KOENIG, AMY L . ; and VANDEGEEST,
KiMBESLY A. Inhibitory Control in Young Children and Its Role in Emerging IntemaHzation. GHILD DEVELOPMENT, 1996,67,490-507. We examined inhibitory control as a quality of temperament that contributes to internalization. Children were assessed twice, at 26-41 months (N = 103) and at 43-56 months (N = .99), on repeated occasions, in multiple observational contexts and using parental reports. Comprehensive behavioral batteries incorporating multiple tasks were designed to measure inhibitory control at toddler and preschool age. They had good internal consistencies, corresponded with maternal ratings, and were developmentally sensitive. Individual children's performance was significantly correlated across both assessments, indicating stable individual differences. Girls surpassed boys at both ages. Children's internalization was observed while they were alone wiih prohibited objects, with a mundane chore, playing games that occasioned cheating, being induced to violate standards of conduct, and assessed using maternal reports. Inhibitory control was significantly associated with internalization, both contemporaneously and as a predictor in the longitudinal sense. The implications for considering children's temperament as a significant, yet often neglected contributor to developing internalization are discussed.
Conscience and internalization, the central goals of socialization, have recently resurfaced in developmental research. The review of extant literature (Kochanska, 1993) leads to a conclusion that the two predominant explanations why children differ in outcomes of intemahzation involve either differences in their socialization histories (psychodynamic, social-learning, attributionai traditions) or in their level of understanding rules and social perspectives (cognitive-developmental model). Recently, Grusee and Goodnow (1994) bridged socialization and information-processing approaches in a major theoretical effort. Surprisingly, until recently, the role of child temperament in internalization has been virtually ignored. In the last few years, however, research about its importance in early moral development has begun to con-
verge theoretically and empirically (Asendorpf fie Nunner-Winkler, 1992; Eisenberg & Fabes, 1992; Kochanska, 1993; Pulklcinen, 1986; Rothbart & Ahadi, 1994; Rothbart. Ahadi, & Hershey, 1994). Although difficult temperament has been linked to undercontro! (Bates, 1989a; Lytton, 1^0), temperamental qualities tfiat foster successful socialization are poorly understood. We propose that two temperamental inhibitory systems, passive and active, may be implicated as important facilitators of conscience development (Kochanska, 1993). Rothbart (1989a) proposed a useful dis^.^^^^^ between passive and active inhibition (characteristic, respectively, of the reactive and self-regulatory aspects of temperament). The passive inhibition system encompasses anxiety, shyness, fearftilness, and inhibition to the unfamiliar (Kagan,
This research was supported by the grants from the National Science Foundation (DBS9209559), the MacArthur Foundation, and from the University of Iowa (Faculty Scholar Award) to Grazyna Kochanska. The assessments of preschool-age inhibitory control and behavior with the prohibited toys were, respectively, the third and fiMi authors' honors theses. We appreciate Mary Rothbart's expertise on the theory, research, and measurement of temperament and Jerome Kagan's helpful guidance regarding the inhibitory control battery and tfie mischiefs p^adigms. We are very grateful to Juli Fratzke, Kathy DeVet, Margi Goldman, Eric Poole, Nazan Aksan, Samuel Putnam, and many undergraduate students for the help with data collection and coding, and to Lloyd Frei for adapting the pinbali toy for one of our tasks. The Toddler Study families deserve credit for their enthusiastic commitment to the project Please send all correspondence to Grazyna Kochanska, Department of Psychology, University of Iowa, Iowa City, IA 52242.
[Child Development, 1996,67,490-507, 1996 by the Society Cot Research in Child Development, Inc. Al! rights reserved. 0009-3920/96/6702-0016$01.00]
Kochanska et al.
1989). In other papers originating from our current research program, we reported the main and moderating effects of children's fearfulness on the development of intemalization (Kochanska, DeVet, Goldman, Murray, & Putnam, 1994; Kochanska, 1995); others have aiso reported similar findings (Asendorpf & Nunner-Winkler, 1992; Rothhart et al., 1994). The active inhibition system reflects effortful impulse control. This temperament system denotes more active processes of inhibition, effortful or willful control of actions, and self-regulation, capable of regulating both approach and avoidance (Rothbart, 1989b), that may be described using the terms "inhibitory control" or reversed "impulsivity."^ Others also view inhibitory control as an active rather than passive process (Block & Block, 1980; Olson, 1989; Reed, Pien, & Rothbart, 1984; Schachar & Logan, 1990). Already in the nineteenth century, several theorists emphasized the importance of "will" in children's development and its active, deliberate, and executive qualities (Kagan, 1984). Compared to other temperament qualities that have attracted attention, inhibitory control is a particularly difficult research topic. In contrast to temperament traits such as, for instance, fearfulness to the unfamiliar or distress to limitations, that may be conceptualized as discrete emotions or reactions to changing stimulation, inhibitory control is a much more complex construct. It is a capacity for active, voluntary inhibition or modulation of conduct, or self-regulation. As other research has shown, emerging selfregulation integrates inputs from several developing systems, including cognitive functions such as memory or language (Kopp, 1982), or the developing attentional mechanisms (Rothbart & Ahadi, 1994; Rothbart, Derryberry, & Posner, 1994). Parent-child interaction factors, including qualities of attachment (Londerville & Main, 1981; Matas, Arend, & Sroufe, 1978; Stayton, Hogan, & Ainsworth, 1971) and parental control (Maccoby, 1980; Olson, Bates, & Bayles, 1990; Power & Chapieski, 1986; Silverman & Ragusa, 1992), have also been identified as con-
491
tributors to the growth of inhibitory and impulse control processes. To complicate matters further, in contrast to temperament traits such as fearfulness or anger, inhibitory control emerges relatively late in development, in late infancy, at the transition into the second year, and develops further in the toddler and preschool years (Rothbart, 1989a). By then, considerable progress in other domains, including cognition, has already occurred, and the child already has a history of meaningful social relationships. Thus, it is often difficult to disentangle; several related processes that all have contributed to self-regulation: temperament, history of parent-child relationship and discipline, and cognitive maturation. Without questioning the importance of socialization and cognitive factors, it is reasonable to assume that children's temperament, in part biologically based, may be a significant contributor to children's emerging self-regulation. Recently, Rothbart and colleagues (Rothbart et al., 1994) described the biological underpinnings of the dimension of effortful control. In particular, they focused on the structures in the midprefrontal cortex (the anterior attention network) as involved in the effortful control of behavior in childhood and the dimension of constraint in adulthood. Other research also found biological correlates of impulsivity (Harburg, Gleibermann, Gershowitz, Ozgoren, & Kulik, 1982). The multifaceted origins and complexity of inhibitory control should not preclude research efforts, however. This quality is strongly implicated in socialization, which often requires that children actively inhibit their impulses and comply with caregivers' standards of conduct Consequently, in the conceptual model guiding our work (Kochanska, 1993), we postulated that inhibitory control may be one ofthe individual characteristics that make children better or worse prepared to internalize family and society standards, and thus it may be a temperamental contribution to the complex process whose final outcome depends on multiple factors, including parents' and peers' influences, cognitive maturation, and social cog-
^ The relations between inhibitory control and impulsivity are more complex than the latter being simply the reversed former; inhibitory control is defined as the capacity to plan and to suppress inappropriate approach responses under instruction, and impulsivity as the speed of response initiation (Rothbart et al., 1994), and they, in fact, load on different broad temperament factors. Many tasks, however, appear to assess both qualities, and there indeed is a significant overlap between the two. Thus, both were used in this study's behavioral batteries and maternal ratings.
492
Child Development
Recently, there have been advances m observational temperament batteries, but mostly limited to reactive aspects of temperament, such as fear of the unfamiliar (RezResearch supports such a contention. nick, Gibbons, Johnson, & McDonough, Milich and Kramer (1984) reviewed evi- 1989) or discrete emotions, such as ^ g e r , dence on the relations between low impul- fear, or joy (LAB-TAB; Goldsmith & Rothsivity and resistance to temptation. Block bart, 1992). There is no analogous behavioral and Block (1980) examined ego-control and battery to assess inhibitory control in young ego-resilience, concepts that include such children. Therefore, a separate important functions as delay of gratification or motor step in this work that preceded the actual inhibition, and thus closely relate conceptu- investigation ofthe links between inhibitory ally and empirically to inhibitory control control and intemalization was to d e s i ^ two (Rothbart & Ahadi, 1994). In a recent longi- such batteries, one for the toddler and one tudinal study. Block and colleagues found for preschool age. We constructed comprethat preschool-age low ego-control and ego- hensive batteries, so that by tapping multiresilience predict behavior indicative of low ple inhibitory control functions, they prointemalization (drug use) in adolescence vided a robust assessment of individual (Block et al., 1988). Finally, in a significant differences in young children. We examined paper, Rothbart postulated links between their intemal consistency, developmental children's inhibitory control and their devel- stability, and validity, the latter reflected by opment of several dimensions of morality the relations with maternal reports. Using (Rothbart & Ahadi, 1994). Empirical support both laboratory observations and parentjj for those links, using parent-reported and ratings combines the advantages of both behavioral measures of young children's ef- methods while remedying the biases inherfortful control (Rothbart et al., 1994), was ob- ent in an exclusive use of either approach tained. In a preliminary study, we also found (Bates, 1989b; Kagan, 1992). strong associations between children's inhibitory control and their development of To examine inhibitory control as both a conscience, although a serious limitation contemporaneous correlate of intemalizawas the exclusive reliance on parental re- tion and its predictor in-a longitudinal sense, ports (Kochanska et al, 1994). The main goal we adopted a cross-sectional/longitudinal of the present work was to explore further design. We assessed temperament and inthe role of inhibitory control in the process temalization at toddler age (Time 1) and at of emerging intemalization, using behav- preschool age (Time 2). We used multiple ioral as well as mother-reported measures. paradigms and measures to assess intemalization. In most paradigms in this reConceptual complexities notwithstand- port, children were observed while they ing, another reason why so little is known thought they were without surveillance. At about inhibitory control in childhood in gen- both times, we observed children's intemaleral and its role in emerging intemalization ization of mothers' prohibitions requiring in particular is the absence of widely ac- refraining from an attractive activity. Addicepted, reliable assessments. Researchers tionally, at Time 2, we exMnined intemalizahave used cognitive tempo tasks (Kagan, tion of mothers* requests requiring the susRosman, Day, Albert, & Phillips, 1964), taining of a mundane activity. At Time 2, we slow-down tasks (Maccoby, Dowley, Hagen, also assessed children's intemalization of & Degerman, 1965; Milich & Kramer, 1984; the experimenter's rules while playing Olson, 1989), delay tasks (Campbell, Szu- games that occasioned cheating. Finally, mowski, Ewing, Gluck, & Breaux, 1982; also at Time 2, we observed Aeir reactions Olson et aL, 1990; Vaughn, Kopp, & Krakow, to being induced by tiie exjwrinwnter to 1984), or "go-no go" paradigms (Reed et break established rules of conduct. At both ai., 1984), but the procedures varied de- times, we obtained mothers' reports of chilpending on the investigator. Additionally, dren's intemalization. with a few exceptions (Campbell et al., 1982; We took mediodological precautions to Silverman & Ragusa, 1992; Vaughn et al., 1984), young toddlers have rarely been minimize the unavoidable overlap between the measures of temperament and those of studied. intemalization. Although both sets of paraThe improvement of objective assess- digms involve voluntary control of behavior ment is an important condition of progress and obedience of adults' rules, we t e m p t e d in research on temperament (Kagan, 1992). to sharpen the difierences between them. nition (Block, Block, & Keyes, 1988; Brody & Shaffer, 1982; Grusec & Goodnow, 1994; Rothbart & Ahadi, 1994).
Kochanska et al.
The inhibitory control tasks were never phrased in terms of prohibitions, but rather were presented as challenging tests of skill or games ("let's see if you can . . . "). The experimenter never communicated to the child that his or her performance was substandard. If a task involved competing with the experimenter, for example regarding a delay, she terminated her performance as soon as the child did. In contrast, intemalization paradigms involved situations when the child was facing a temptation to violate explicit standards of behavior. Those stan. dards were conveyed either by prohibitions and requests by an adult (mother or experimenter), in some cases tagged with a clear moral qualification (cheating games), or they involved common standards that had surely been conveyed in past socialization ("mischiefs"). For further discussion of this methodological issue, see Kochanska (1993). This report originates from a large study on socialization and temperament in the early development of conscience. More data on the sample, design, and other measures may be found, for example, in Kochanska (1995), Kochanska and Aksan (1995), or Kochanska, Aksan, and Koenig (1995). All behavioral data on inhibitory control and internalization were based on observations of children on repeated occasions, in multiple naturalistic, yet standard contexts. We place a strong emphasis on the robust measurement; therefore, the measures were aggregated whenever appropriate (Epstein, 1986; Rushton, Brainerd, & Pressley, 1983).
493
children. The rooms are separated by a folding door. All sessions were videotaped. Except for a few tasks that were coded live, all behavioral data were coded from videotapes by separate teams of coders.
SAMPLE
At Time 1,103 normally developing toddlers (51 girls, 52 boys), aged 26 to 41 months (M = 32.86, SD - 4.09), and their mothers volunteered in response to ads in the community. The mothers represented a range of education (12% high school only, 31% postgraduate education) and income (13% under $15,000, 38% over $45,000). Eighty percent were Caucasian, 7% minorities, and 13% did not report race. At Time 2, 99 mothers and children (49 girls, 50 boys) retumed when children were 43-56 months old (M = 46.01, SD = 2.62). Two families had moved, one could not be found, and one refused due to time constraints.
INHIBITORY CONTROL MEASURES
Method
OVERVIEW OF THE STUDY
Mothers and children participated at Time 1, when children were toddlers, and at Time 2, at preschool age. At Time 1, data were collected during two 2-3-hour sessions, one at home and one in the university laboratory (typically, 1-3 weeks later), both conducted by the same experimenter, who had established friendly rapport with the mother and child. At Time 2, there was one 3-3V2 hour laboratory session, conducted by a new experimenter. She administered all inhibitory control tasks. The laboratory included a playroom and a "living room," naturalistically fumished, with a low shelf containing objects extremely attractive to young
Behavioral Observations: The Multi-Task Batteries Two batteries were designed, one for Time 1 and one for Time 2. Each took cumulatively about 50 min, but the tasks were interspersed with other activities. The mother was usually in the room, busy with questionnaires. The tasks were based on a review of past research (e.g., Campbell et al., 1982; Maccoby et al., 1965; Milich & Kramer, 1984; Olson et al., 1990; Reed et al., 1984) and on the analysis ofthe content of temperament concepts and questionnaire items. The tasks assessed delaying, slowing down motor activity, suppressing/initiating activity to signal (go-no go), lowering voice, and, at Time 2, cognitive reflectivity. Each task involved multiple trials. Typically, midway through each task, the experimenter reminded the child ofthe relevant rule, independent of the child's performance. For each task, the final inhibitory control score was the aggregate of all trials. Time 1 battery.--At Time 1, the battery consisted of seven^ tasks, five administered at home and two in the laboratory. Delaying was assessed in four tasks. In Snack Delay, the child had to wait for the experimenter to ring a bell before retrieving
The actual battery consisted of 10 tasks. Two tasks, however, did not appear to correlate with any others (a persistence task involving an inaccessible toy and a task measuring the latency to make a decision), and one "go-no go" task was too hard to understand for 30% ofthe children.
494
Child Development
Suppressing/initiating activity to signal was assessed in Tower. The child was asked to take turns with the experimenter while building a tower using 20 blocks (three trials). The experimenter deliberately waited before placing each block until the child spontaneously signaled that he or she was giving her a tum. Goding of each trial refiected the proportion of blocks placed by the child in the total number of blocks. The score was reversed so that higher scores denoted higher inhibitory control; thus, a score of a child who gave the experimenter all due turns would equal one-half of the total blocks. An additional credit was given for removing blocks carefully to prevent the tower from falling, and a penalty subtracted for pushing the tower over prematurely. The mean score was used. Lowering voice was measured in the Whisper task. The child was asked to whisper the names of 10 consecutively presented cartoon characters. Codes ranged from 0 to 3 (0 = shout, 1 = normal tone, 2 = no response, 3 = whisper); the mean score was used. Reliability of coding.--Except for Whisper, coded live by an observer and reviewed by the main coder, all tasks were coded from the videotapes; 14 entire cases were used for reliability. The kappas were 1.0 for Snack Delay, Turtle-and-Rabbit path accuracy scores, and Lab Gift peeking score, and .88 for Home Gift peeking score. For the latency and the Tower scores, Gronbach's alphas were used (Bakeman & Gottman, 1989, pp. 92-96, generalizability). The alphas were 1.00 for the Tongue and Tower scores and for the Turtle in Turtle-and-Rabbit, .93 for baseline and .96 for the Rabbit in Turtleand-Rabbit, .99 for the latency to adjust the blindfold, .95 for the latency to remove it (Home Gift), 1.00 for the latency to peek over the shoulder, and .97 for the latency to turn around (Lab Gifl). Time 2 battery.--At Time 2, the battery included 12 tasks.^ Five tasks were essentially analogous in terms of instruction and coding to those at Time 1:^ Snack Delay, Whisper, Tongue, Tower, and Lab Gift. Seven new tasks, appropriate to the older age, were
an M&M from under a glass cup (four trials, delays of 10, 20, 30, and 15 sec). Halfway through the delay, ihe experimenter lifted the bell but did not ring it. Coding ranged from 0 to 4 (0 = eats the M&M before the bell is lifted, 1 = eats the M&M after the bell is lifted, 2 = touches the bell or cup before the bell is lifted, 3 = touches the bell or cup after the bell is lifted, and 4 = waits for the bell to ring before touching cup or bell). The mean score was used. Tongue required that the child hold an M&M directly on his or her tongue (four trials, delays of 10, 20, 30, and 15 sec). Coding reflected the length of delay (in seconds) before the child ate the candy; the mean delay score was used. Home Gift called for delay of gratification. The child, wearing a blindfold, sat in front of the experimenter while she noisily wrapped a gift for the child (1 min). Coding involved (a) the peeking score (0 = removing blindfold, 1 = adjusting blindfold to peek, 2 = no attempt to touch/adjust blindfold, (b) the latency to adjust, and (c) the latency to remove blindfold (60 for a full delay). All three scores correlated highly (rs .83-.90), and thus were standardized and combined into one Home Gift score. Lab Gift differed from Home Gift only in that the child sat on a chair facing away from the table where the experimenter was wrapping the gift, but with no blindfold. Coding involved (a) the peeking score (0 = turning around to peek, 1 = peeking over the shoulder, and 2 = no attempt to peek, (b) the latency to peek over the shoulder, and (c) the latency to turn around to peek (60 for a full delay). Ali three scores were correlated (rs .78-.90), and thus were standardized and aggregated into one Lab Gift score. Slowing down motor activity was measured in Turtle-and-Rabbit. The child moved a same-sex doll (baseline), a fast rabbit, and a slow turtle from the start to the finish (Fisher Price bam) along a curving path through a "meadow," two trials for each. Coding involved accuracy in negotiating the path (0 = off path, 1 = occasional departures from path, 2 = essentially within = path), and the slow-down score, the difference between the means of the turtle and rabbit trials.
^ The actual battery involved 13 tasks, but one, that assessed decision time to choose a prize, was dropped because it lowered the overall reliability. We made some minor changes at Time 2. The numbers of trials in the Tower and Tongue * tasks were abbreviated; to adjust the challenge posed by the Snack task to the older age, we introduced an additional demand (child was asked to keep hands on the table during the delay), refiected in the change in the coding (range 0 to 10 for each trial).
Kochanska et al.
added, either to expand the battery or substitute the toddler-age tasks. Delaying included one new task. Dinky Toys. After one of the activities, the child chose one prize from a plastic box filled with small toys (party favors, pencils, stickers, etc.). He or she was asked to put hands flat on his or her knees while the experimenter showed the child the contents of the box, holding them immobile while deciding on the toy, and then to tell the experimenter (without pointing) which toy he or she wanted (for up to 2 min). Coding ranged from 0 to 5 (0 = grabs a toy, 1 = touches a toy but does not take it out, 2 = points to toys without touching, 3 = removes hands from knees but does not point, 4 = hands twitching but remaining on the knees, 5 = hands immobile on the knees). Slowing down motor activity.--There were three new tasks. In Walk-a-LineSlowly, adapted from Maccoby et al. (1965), the child was asked to walk down the "path" (a 2V2 inch x 12 foot strip of floral fabric taped to the floor; there were a baseline trial and two slow trials). Coding involved the duration (in seconds) of each trial. The score was the mean ofthe two slow trials. In Telephone Poles (adapted from Draw-a-LineSlowly; Maccoby et al., 1965), the child drew the "wires for the squirrels to play on" between two telephone poles 10 inches apart for three trials (baseline, as fast as possible, and as slow as possible). Coding involved the dnration of each trial (in seconds). The score was the difference between the slow and the fast trials. Circle, similar to the Telephone Poles, required that the child draw "a round path for the turtle to walk on," 5% inches in diameter (baseline and twice as slow as possible). Goding involved the duration, in seconds. The score was the mean ofthe two slow trials. The two drawing scores. Telephone Poles and Circle, correlated (r = .44, p < .001), and thus were aggregated into one Drawing score. Suppressing/initiating activity to signal included two new tasks. Bear and Dragon was adapted from Reed et al. (1984). The experimenter used two hand puppets, a Bear and a Dragon, that issued directives to the child commanding various movements (e.g., "Touch your tummy"). The child was asked to perform the movement requested by the Bear, but to inhibit the one requested by the Dragon (six trials for each puppet). Coding
495
for each Bear trial was 0 = failure to move, 1 = a wrong movement, 2 = a partial movement, 3 = a full, correct movement; for each Dragon trial 0 = a full movement, 1 = a wrong movement, 2 = a partial movement, and 3 = a full inhibition of movement. Two scores were retained, an average for the Bear trials and an average for the Dragon trials. Pinball was also adapted from Reed et iil. (1984), and measured the child's ability to (a) suppress an act until a signal (the first six trials), and (b) to suppress an act in response to one signal and perform an act in response to another signal (the next six trials). A big pinball toy was adapted and used for this paradigm. The first six trials required the child to hold the plunger, connected to the sensor, in the pulled out position until tlie experimenter said "Go!" (delays of 10, 15, 25, 15, 20, 10 sec). The next six trials called for the child to release the plunger when the experimenter showed a green sign and to continue holding the plunger when the red sign was shown. When the plunger was first pulled, a sound began; the sound stopped when the sensor detected that the child released the plunger by 2 mm. Three red and three green signs were shown, in mixed but standard order, 10 sec apart. Coding assigned 1 point for each 5 sec of successful delay, and subtracted 1 point if the child began to release the plunger (so that the sound stopped even if the ball did not move) prematurely or to the red sign. Two scores were derived (the aggregates for the first and the second sets of trials). Because they were correlated (r - .68, p < .001), they were aggregated into one Pinball score. Cognitive reflectivity was captured by KRISP (Wright, 1971). This measure of impulsivity/reflectivity calls for matching a picture to one in a set where all but one differ by minor details. Coding involved the average latency to respond^ and the correctness score (reversed total number of errors). Reliability of coding.--Two coders established reliability on 15% of the sample. The kappas were used for categorical scores (.97 for Snack Delay, 1.00 for Dinky Toys, and .89 for the Lab Gift peeking score). Alphas were used for the continuous scores aggregated across trials (Bakeman & Gottman, 1989, pp. 92-96). The alphas were 1.00 for Tower, .99 for Bear, 1.00 for Dragon, 1.00 for Tongue, .99 for Pinball, and .99 for Lab Gift latencies. Walk-a-Line-Slowly was conducted in the hallway; the experimenter
The latency score in KRISP lowered the reliability and was dropped.
496
Child Development
high competing activity (at Time 2, also of low gentle touch). Internalization of maternal request without surveillance (Alone with a Mundane Task: Time 2).--Following a 10-miu mother-child toy cleanup, mother asked the child to complete that task by him- or herself. The child was then left alone. Every 10-sec segment was coded. The code of internalized cleanup was further used (the child engaged in cleaning up and no play behavior present). The N for this variable is 85, because 14 children had completed the cleanup before the mother's departure. Reliability, kappa, was .86. Internalization of the experimenter's prohibition (Alone with the Cheating Games: Time 2).--At three points during the session, the child played "special games with great prizes" in the playroom, while mother was in the living room. The experimenter instructed the child about each game's rules, added a serious comment on the importance of not cheating, and promised prizes from a large bin with attractively wrapped packages for winning. Each game was practically impossible to win if the rules were fojiowed. Each time, the experimenter "remembered" that she had to leave to go to the other room for a while, and the child was left for 3 min to play the game alone. The experimenter warned the child before reentering the room, and then "discovered" that the game had been set up wrong. Each child then played the game again and won a prize. In the Animal Game, the child tried to identify three animals hidden under cloth, without looking, only by touching with a tip of one finger. In the Bird Game, the child tried to select "magic birds" (marked by stickers on the bottoms) irom among 30 colorful birds, being allowed to touch and examine no more than two. In the Dart Game, the child threw five darts into a ring placed on the floor very far away, not being allowed to leave a narrow space outlined with wooden boards or to throw a dart more than once. We coded the latencies to transgress (99% of codes within 1 sec) and behaviors for each 3-sec segment while the child was alone (kappas .85-.92). The variables, coded
measured the duration of each trial using a stopwatch. The duration of drawing in Telephone Poles and Gircle was measured live using a stopwatch, the fonner by an observer through a one-way mirror, and the latter by the experimenter. Maternal Ratings Time 1 and Time 2.--At both times, mothers filled out Rothbart's Ghildren's Behavior Questionnaire, a weil-estabJished parental instmment (Rothbart et al., 1994) consisting of 15 scales, assessing dimensions of child temperament, with response format from 1 (extremely untrue) to 7 (extremely true). Two scales were selected for this report, inhibitory control (alpha at Time 1, .78, at Time 2, .77), with items such as "can lower his/her voice when asked to do so," and impulsivity (alpha at Time 1, .81, at Time 2, .82), with items such as "usually rushes into an activity without thinking about it."
INTERNALIZATION MEASURES
Most intemalization measures have been described in much detail, including information about reliability and data reduction, in other published work; Alone with Prohibited Toys and mother ratings (Time 1) in Kochanska and Aksan (1995), and Alone with Prohibited Toys, the unsupervised toy cleanup, the "cheating games," and mother ratings (Time 2) in Kochanska et al. (1995). Thus, below we include only brief information about these measures. Internalization of maternal prohibition without surveillance (Alone with Prohibited Toys: Time 1 and Time 2).--Approximately IV2 hours into the session, the child was left alone in the "living room" with the shelf with objects that mother had prohibited him or her from touching upon entry to the lab, and asked to do a dull sorting task. The paradigm lasted 12 min; each 5-sec segment was coded using one of six codes (looking with no attempt to touch, self-correction, gentle touch, deviation, competing activity, and busy with the sorting task). Reliability kappas were above .90; separate agreements, 60%-100%. At both times, the principal components analysis produced an analogous factor,^ Behavioral Intemalization. At both times, it was comprised of low deviation, high looking with no attempt to touch, and
* At Time 2, the PCA also produced another factor, Task-Oriented, mostly composed ot the " child's involvement in tbe sorting task (Kochanska et al., 1995). Because at Time 1, however, we had used only the Behavioral Intemalization factor (Kochanska & Aksan, 1995), for consistencv's sake we decided to use only this factor, analogous at both times, in this report.
Kochanska et al.
similarly in all games, were standardized and aggregated. The latencies to lift the cloth, to touch and examine the first "illegal" bird, put one foot or whole body, or to step away from the marked space were combined into the latency to transgress (alpha = .78). All instances of improperly touching the animals, lifting cloth, peeking under cloth, "illegal" bird examinations, moving the ring closer, being outside the marked space, as well the number of animals seen, darts retrieved, and darts manually put in ring were combined into the extent of transgression (alpha = .76). All instances of looking at animals and birds without touching, touching animals correctly, and being within designated space were combined into rulecompatible conduct (alpha - .55). All three aggregated categories (the extent of transgression reversed) were then combined into one internalization composite (alpha = .83). The reluctance to violate standards of conduct (Mischiefs: Time 2).--Toward the end ofthe session, when the mother was in the living room, the experimenter and the child went to the playroom, where five "mischiefs" took place. In each, E first asked child to perform a "legal" act and then an analogous "illegal" act, violating an established standard of conduct (Kagan, personal communication, The 1993). directives were issued in a friendly, matter-of-fact fashion. Up to two neutrally delivered repetitions of the request were given, if necessary. 1) Scribble: The child was asked to scribble on a drawing pad with a black marker (legal act), and on a page of a book the experimenter said she "used for school" (illegal act). (2) Tear: The child was asked to tear the scribbled page out ofthe pad (legal act), and out ofthe textbook (illegal act). (3) Rail: The child was asked to throw a nerf ball at the wall (legal act), and at the experimenter's face (illegal act). (4) Spill: An empty cup and one with a standard amount of water were placed on the table. The child was asked to pour the water into the empty cup (legal act), and on the floor (illegal act). (5) Photo: The child was asked to tear a paper card to pieces (legal act), and to tear up a same-size photo ofthe experimenter (illegal act). The experimenter then thoroughly debriefed the child. She said, "these were some silly things they did," and explained that the textbook was old and used only for scribbling, that there was a stain on the floor that needed to be washed away with the wa-
497
ter, and that the photo had a spot on it and she had another, good one (which she produced). She also stressed that "one would never do things like that for real." No child or mother was distressed after the paradigm; the mothers had been informed ahead of time, and none expressed reservations. For each act, we coded latency to perform it (in seconds), the readiness to perform it (0 = never, 1 ^ after two additional prompts, 2 - after one additional prompt, 3 = after initial directive), and lack of reluctance (1 - child appears very uncomfortable, argues/questions/refuses the directive, shifts in chair, bites lip; 2 = hesitates, looks back and forth at experimenter, appears somewhat reluctant; 3 = performs the act with no visible hesitation, readily, or gleefully). If the child never performed the act, that is, readiness = 0 (no children in Scribble, one in Tear, four in Ball, five in Spill, three in Photo), latency was not coded, but the lack of reluctance was coded. Ten cases (100 acts) were used for reliability. The kappas were .90 for readiness and .84 for lack of reluctance. For reliability, the latencies were coded with a greater accuracy (up to one one-hundredth of a second); 98% of all judgments were within 1 sec, 100% were within 1.50 sec. In data aggregation, to create a robust measure of reluctance to violate the established standards of conduct, we aggregated the latencies and the scores on readiness and on the lack of reluctance to perform an act, separately for the legal and illegal acts. The average latency to illegal acts (M = 4.28, SD = 2.60) was longer than that to legal acts (M - 3.26, SD = 1.61), ( = -3.98, p < .001. The readiness to illegal acts (M = 2.72, SD = .38) was less than to legal acts (M = 2.93, SD = .14), ( = 5.52, p < .005, and so was the lack of reluctance (illegal, M = 2.56, SD - .32; legal, M = 2.92, SD = .15), t = 11.74, p < .001. Thus, for each child, we created difference scores for the average latency (illegal minus legal), readiness (legal minus illegal), and for the lack of reluctance (legal minus illegal), standardized them, and aggregated. That composite refiected the degree of relative reluctance to violate standards (alpha = .80); higher scores reflected higher relative reluctance to violate standards of conduct. Maternal ratings.--At both Time 1 and 2, mothers rated their children on a 20-item internalization scale (1 = extremely untrue, 7 = extremely true), part of a larger con-
498
Child Development
p < .025), and for boys (r = .66, p < .001), for the composite scores. There was also stability across both assessments in matemal ratings; for inhibitory control, r = .67, p < .001 (girls .76, boys .58), and for impulsivity, r = .69, p < .001 (girls .65, boys .70). Correspondence between Behavioral and Mother-Reported Inhibitory Control The relations between children's scores in the behavioral batteries and maternal ratings indicated a meaningful convergence of the behaviors and mother reports. At Time 1, the battery correlated .30, p < .001 with mother-reported inhibitory control, and - .26, p < .005 with impulsivity. At Time 2, the correlations were .42, p < .001 and - .26, p < .005 (covarying current age). Developmental and Gender Effects in Inhibitory Control We first examined developmental changes using correlations. At Time 1, for the whole battery composite, children's performance increased between 26 and 41 months (r = .56, p < .001). Similarly, mothers of older children rated them as higher on inhibitory control than mothers of younger children (r = .30, p < .001), although it was not true for impulsivity (r ^ - .05). At Time 2, children's performance in the battery also improved wi5i age (r = .31, p < .(X)l). Mothers of older children again rated them as higher on inhibitory control (r = .26, p < .005), but the correlation for impulsivity was marginal (r = -.18). Longitudinal comparisons: Time 1 versus Time 2.--Next, we examined longitudinal changes (Time 1 vs. Time 2) in children's performance, using the subset of four tasks identical at both times (Whisper, Tongue, Tower, and Lab Gift). Lab Gift was decomposed to its three components (Time 1 first. Time 2 second): the peeking score (M = .46, SD = .80, M = 1.03, SD = .93), the latency to peek over shoulder (M = 17.17, SD = 22.73, M = 37.33, SD = 23.60), and the latency to tum around to peek (M = 26.03, SD = 24.73, M = 44.99, SD = 20.85). These measures, before standardization (for each, the scores at Time 1 and 2), were the dependent variables in a MANOVA; type of task and time of assessment were the withinsubject factors, and child gender the between-subjecis factor. Not surprisingly, children's performance improved between Time 1 and Time 2, as indicated by a significant interaction of type of task and time of assessment, F(5, 93) = 75.82, p < .001.
science instrument (Kochanska et al., 1994). It describes intemalization of family rules, compliance when alone, etc. (Time 1 alpha .86, Time 2, .88). The ratings correlated across Time 1 and 2 (r = .68, p < .001). Table 1 presents the overview of the Time 1 and T'ime 2 measures. The means and standard deviations, for the entire sample and by gender, are included.
Results
Overview The analyses encompassed three steps. First, we explored the internal consistencies and developmental continuity of the newly developed inhibitory control batteries. Second, we examined iheir validity by exploring the correspondence between maternal reports and chiidren's observed performance in the batteries, as well as the developmental and gender differences in children's inhibitory control. Third, we examined the associations between inhibitory control and intemalization, using observed and mother-rated measures, both contemporaneous at toddler and preschool ages, and longitudinal, where toddler-age inhibitory control was considered the predictor of intemalization at preschool age. Internal Consistency of Inhibitory Control Batteries At Time 1, following the inspection of the intercorrelations among the tasks, we standardized all the scores and examined their internal eonsistency. The tasks clearly tapped a common capacity, resulting in alpha = .75. The average item-total correlation was .45. Also, at Time 2, we inspected the intercorrelations among the tasks and standardized the raw scores. Again, the battery appeared internally consistent, with alpha = .79, and the average item-total correlation .42. In view of these findings, we created two composite scores, for Time 1 and for Time 2, to represent chiid overall inhibitory control (the mean ofthe standardized scores on all tasks), and used them in all following analyses except those focused on longitudinal comparisons of performance, conducted for the subset of tasks similar at both times. Longitudinal Stability in Children's Inhibitory Control There was considerable longitudinal stability in children's overall performance in the batteries at toddler and preschool ages, r = .59, p < .001, both for girls (r = .35,
Kochanska el al.
The paired ( tests, all significant at p < .001, revealed improvement in each task (see scores for the entire group. Table 1). For mothers' ratings, we conducted a similar MANOVA, with the scores on inhibitory control and impulsivity at Time 1 and 2 as the dependent variables. The interaction of the trait and the time of assessment was significant, F(l, 97) = 14.24, p < .001. The subsequent paired t tests indicated that mothers rated child inhibitory controi as higher at Time 2 (t = -4.44, p < .001), although there was no difference in rated impulsivity it = 1.47). Data are in Table 1. Gender comparisons.--We conducted two MANOVAs (one with the behavioral batteries composite, and one for motherreported scores as the dependent variables), with child gender as the between-subjects factor. Both Time 1 and Time 2 measures were entered; the time of assessment was the within-subject factor. Effect of sex was significant in both analyses; for the behavioral measures, F(l, 97) = 17.79, p < .001, and for mothers' reports, F(l, 97) = 7.94, p < .01. For the former analysis, ( tests revealed that girls were higher at Time 1 (* = 2.89, p < .005) and at Time 2 (f = 4.50, p < .001). For the latter analysis, girls were rated higher than boys in inhibitory control (Time 1, t = 2.02, p < .05; Time 2, t = 2.79, p < .025), and lower in impulsivity (Time l,t = - 2 . 1 0 , p < . 0 5 ; T i m e 2 , t = -2.31, p < .05). The means are in Table 1. Contribution of Inhibitory Control to Children's Intemalization We conducted a series of correlations between behavioral (composite) inhibitory control scores and mother-rated inhibitory control and impulsivity and children's internalization measures. All significant associations were in the predicted direction. The data are in Table 2. Contemporaneous associations.--At Time 1, compared to children who had low scores in the inhibitory control battery, those with high scores when alone with prohibited toys were more internalized, and were seen by their mothers as more internalized in their daily lives. Compared to children rated by mothers as low on inhibitory control and high on impulsivity, those rated as high on inhibitory control and low on impulsivity were more internalized when alone with the toys, and were rated as more internalized in daily lives. At Time 2, again there were many robust links. Children with high
499
scores in the inhibitory control battery were more internalized when alone with prohibited toys, while alone with the cleanup, and while playing the games occasioning cheating. They were reluctant to break established rules of conduct while being induced to do so, although this correlation was low. They were again rated as more internalized in daily lives. Similarly, children seen by their mothers as high on inhibitory control and low on impulsivity were highly internalized in all observational paradigms except the cleanup and received matemal high ratings of intemalization in daily life. Longitudinal associations.--There were significant associations between toddler-age inhibitory control and preschool-age internalization. Toddlers who had high scores on the behavioral battery as preschoolers were more internalized while alone with the forbidden toys, with the cleanup, and with the cheating games, and were seen by the mothers as highly internalized. Those who as toddlers had been rated by mothers as high on inhibitory control and low on impulsivity, at age 4 were more internalized in most measures. Multivariate hngitttdinal analyses.--The final analyses, multivariate hierarchical regressions, examined the unique contributions of Time 1 and Time 2 inhibitory control to Time 2 intemalization, and, specifically, explored if Time 2 inhibitory control explained additional variance in internalization beyond that explained by Time 1 inhibitory control. To gain a more general view of the relations than that produced by the correlations, we created an overall Time 2 intemalization score by averaging each child's scores across paradigms: Behavioral Intemalization (prohibited toys), internalized cleanup (mundane task; standardized), the cheating games' composite, the relative reluctance to violate standards, and maternal rating (standardized). That score, representing a multifaceted "conscience" (alpha .55), was the dependent variable in all equations. We conducted two separate regressions; in one, the inhibitory control batteries' composites, and in the other, maternal ratings of inhibitory control and impulsivity were the predictors. To control for child sex and ages at Time 1 and 2, these variables were entered at Step 1 in each equation. This step explained 16% of variance, F^i^(3,95) = 5.88, p < .005. Following the developmental se-
I
[
Q
tfl in
o
en
CO 00 t o U5 CD ( O t o T
1
0 II CQ
_,, Q
Tl" ;
o
Q
Hi
!O op Q G O O ID I O to
t-. o t-:
-H M r>i
S
CO --" CO O
01 0
Q
C Q
in 1 )
in
ol
Q
s s
"o "o -2 ^ S n c/^ o
Ta-o .S g
,Q V
S5 o p 22
QQ
00 be
en aa a.
500
501
I
o I
be
E H
S CO
o
II
; S
c o s
o
1, (N
Kochanska et al.
quence, Time 1 data were then entered at Step 2 (the inhibitory control battery score in the first, mothers' ratings in the second equation), and Time 2 data were entered at Step 3 (the inhibitory control battery score in the first, mothers' ratings in the second equation). In the first equation, Step 2 (Time 1 inhibitory control battery score) added 4% of variance, F<,h(4,94) = 5.27, p < .025, beta = .28. Time 2 battery score (Step 3) explained an additional 10% of variance, Fch(5, 93) = 13.58, p < .001, beta = .42. The final R^ = .30, F(5, 93) = 8.09, p < .0001. In the second equation, mothers' ratings of inhibitory control and impulsivity at Time 1 (Step 2) added 19% of explained variance, F,h(5, 93) = 13.58, p < .001, and the ratings at Time 2 (Step 3) added 9%, F^^C?, 91) = 7.25, p < .005. The final R^ = .43, F(7, 91) = 10.09, p < .0001. To summarize, both univariate and multivariate analyses were consistent. Together, they supported the view that inhibitory control at toddler and preschool age makes significant contributions to a general internalized "conscience," assessed at preschool
503
ment quality that underpins children's performance on multiple tasks that call for different functions (delaying, the slowing down of gross and fine motor activity, the initiating and inhibiting of response to signal, the lowering ofthe voice, reflective information processing). High stability over time is particularly noteworthy. The batteries consist of tasks that may be easily replicated by others, are sensitive to developmental changes, and converge significantly with matemal reports. Thus, they continue and extend progress in temperament assessment beyond its reactive aspects (Goldsmith & Rothbart, 1992) into its effortful or self-regulatory dimension, particularly relevant for socialization. In the current work, we aimed at creating batteries that are comprehensive in nature. Even though all tasks share the effortful, willful, and intentional quality, they differ in the specific task demands. In future work, a more fine-grained analysis of possible different skills that constitute the components of general inhibitory control may prove fruitful. Future investigators may wish to pinpoint more specific early temperamental antecedents of broad-ranging developmental outcomes. Compliance with caregivers, prosocial and aggressive conduct, internalized or undersocialized conduct, achievement and mastery, peer relations, coping, regulation of anger, and general social competence have ail been associated with the early development of inhibitory control and self-regulation (Block & Block, 1980; Eisenberg & Fabes, 1992; Eisenberg, Fabes, Nyman, Bemzweig, & Pinuelas, 1994; Maccoby, 1980; Olson, 1989; Puikkinen, 1986; Rothbart & Ahadi, 1994). It is possible, however, that specific inhibitory control skills are particularly important for specific competencies. For example, Olson (1989) found that only some measures of inhibitory control were associated with social development outcomes. In future work, using our multifaceted, internally consistent batteries will allow for a better understand-
Discussion
This work makes several contributions to developmental research. First, this study results in methodological progress in the assessment of inhibitory control. Second, it highlights the important role that this temperament quality plays iu the process of emerging intemalization of rules of conduct. The development ofthe behavioral batteries to assess inhibitory control at toddler and preschool age is a contribution in its own right. We provide investigators with new observational paradigms to study inhibitory control in young children. High internal consistency, at both times, indicates that the batteries indeed tap a common tempera-
^ We also wished to demonstrate that individual ditferences in children's temperamental inhibitory control contribute to the development of conscience, above and beyond the contribution of parental influence. To that efFect, we conducted all analyses ofthe links between inhibitory control and intemalization also controlling for the style of matemal discipline deemphasizing power (two robust composite measures were available, observed and self-reported; see Kochanska, 1995). All correlations remained significant. In the regressions, Time 1 and Time 2 behavioral inhibitory control each added 6% of explained variance in the multifaceted "conscience" score at Time 2 (both significant at p < .01), above and beyond the significant contribution of the two parenting measures that had been entered first. The analogous contributions of Time 1 matemal ratings of inhibitory control and impulsivity were 15% of variance, and Time 2 ratings, an additional 8%, both significant at p < .0025, above and beyond the impact of the parenting measures.
504
Child Development
sures of intemalization, across both times of assessment, and, except for the mischiefs paradigms, were generally robust. Interestingly, the mischiefs differed from other intemalization paradigms in that the child was directly interacting with the experimenter rather than alone with some form of temptation. Thus, in contrast to other paradigms that involved a conflict between a prohibited but attractive act and a less attractive but "moral" act, the mischiefs also may have involved other motivational factors, such as a wish to comply with and please the adult. Despite the significant recent progress in research on individual differences among children and their contributions to social development, there are very few studies of the role of temperament in intemalization. Thus, temperament's relations with the developing conscience remain poorly understood. Researchers focus mostly on environmental factors in the development of capacities essential for successful socialization, such as security of attachment (Londerville & Main, 1981), quality of discipline and control (Hoffman, 1983; Maccoby & Martin, 1983), social cognitive factors involved in discipline (Grusec & Goodnow, 1994), social referencing (Emde, Biringen, Clyman, & Oppenheim, 1991), or mother-child mutually cooperative orientations (Maccoby & Martin, 1983). The current findings do not diminish the importance of socialization, and, indeed, in other papers originating from this research program we focus on parent-child variables as critically important in the development of conscience (Kochanska, 1995; Kochanska & Aksan, 1995; Kochanska et al., 1995). We argue, however, that individual differences in children's inhibitory control, that may, in part, have biological underpinnings, contribute to the development of conscience, above and beyond the contribution of parental influence. The additional analyses fully support this argument (see note 7). Our data thus call for an increased appreciation of children's temperament, so far a neglected perspective in the study of intemalization. In addition, we have recently demonstrated (Kochanska, 1991, 1995) that children's temperament may not only directly contribute to developing intemalization, but that it also may moderate the impact of socialization. In other words, different parental socialization strategies and orientations may be effective in promoting intemaliza-
ing ofthe specific links between this important aspect of children's temperament and their social-emotional d...