RESULTS
Because this study consists of three separate parts and the anticipations for Part 2 and Part 3 depend on the results obtained in Part 1, the Results section necessarily will include some theoretical hypotheses. Part 1: Coherence, Replicability, and Psychological Nature of Preschool Personality Prototypes Determining the Number of Replicable Preschool Personality Prototypes Inverse (Q) factor analysis followed by varimax rotation was employed to identify the personality prototypes. This procedure results in a factor loading for each participant and a (person) factor score for each CCQ item. A prototypical individual loads highly on one factor only (Gorsuch, 1983). Factor loadings were used to measure individual differences; a loading “indexes the degree to which the individual’s particular personality configuration resembles that captured by the factor and thus provides a normed index of membership in each personality prototype category for each individual” (York & John, 1992, p. 495). In contrast, factor scores were used to interpret each personality prototype. Vectors of 100 factor scores (one for each CCQ item) can be correlated and the resulting correlation denotes the degree of factor similarity.
The number of personality prototypes was determined by factor replicability analysis (see Everett, 1983). The sample was randomly divided into two nonoverlapping subsamples and each random half was subjected to inverse factor analysis. The 100 factor scores independently identified in each half were then correlated, the assumption being that only factors identified in each half were replicable. The traditional criterion for acceptable factor replicability is a congruence correlation of .80 or greater (e.g., Asendorpf&van Aken, 1999). Three of our personality prototypes (or person factors) met or
exceeded this criterion and were retained for further analyses.
Fifty-six individuals (55%) received their highest loading on the first personality prototype, 35 (34%) on the second, and 11 (10%) on the third. The three factors explained 65% (Factor 1 = 41%, Factor 2 = 17%, and Factor 3 = 7%) of the total variance in the CCQ evaluations. This factor solution is comparable to the one reported for German children by Asendorpf and van Aken (1999). Consistent with our results, these authors also reported that the factors did not differ for the sexes.2 To ensure that the participants were sufficiently pure representatives of only one personality prototype, they had to load at least .40 on one personality prototype and their second highest loading had to be at least .20 lower than their highest loading. Participants with loadings of .40 higher on more than one prototype were excluded. These criteria successfully classified 83 (39 boys and 44 girls) of the 102 participants, or 84%. Subsequent analyses are based on these 83 participants. To ensure that the three personality prototypes were sufficiently different, the participants were assigned to nonoverlapping groups based on their highest loading (see York & John, 1992). The means and standard deviations of the loadings for each factor were then calculated. As Table 1 shows, the mean factor loadings were substantially higher and the mean standard deviations were substantially lower when the participants were assigned to the prototype on which they had received their highest loading.
Interpreting the Psychological Nature of the Three Personality Prototypes The correlations of the personality prototypes with ego resiliency were .79, .33, and –.33; their correlations with ego undercontrol were –.29, .89, and –.11. A multivariate analysis of variance (MANOVA) showed a significant interaction between the three personality prototypes and the two personality constructs (F = 38.03, p < .001), indicating that the three factors were defined by distinctly different configurations of ego resiliency and ego undercontrol. The three personality prototypes were labeled overcontrolled resilients, resilient undercontrollers, and brittles, respectively. To further evaluate the potential effects of attrition, personality prototypes were developed using all 157 children in the age-4 sample, not only the 102 participating in the current study. Also in this larger sample, three replicable prototypes were discerned. The correlations between these prototypes, based on the partly overlapping samples, were also very high, ranging from .99 to .96. The correlations with the ego resiliency and ego undercontrol measures were also virtually identical for the two sets of prototypes. The 10 most and least characteristic CCQ items are displayed in Table 2 separately for each personality prototype. The first personality prototype, overcontrolled resilient, was characterized by CCQ items such as dependable, attentive, helpful, reasonable, and personable, and not by items indicative of aggression, unworthiness, or stress reactivity. This personality configuration portrays a socially well-adjusted preschooler likely to adapt well to a broad range of situations. Block and Block (1980) noted, “For the overcontrolling child, the presence of ego resiliency results in a high degree of socialization that fits and feels well, a relative absence of anxiety and intimidation in reacting to and acting on the world” (p. 88).
The second personality prototype, resilient undercontrol, was characterized by CCQ items such as energetic, assertive, aggressive, expressive, sociable, and able to stand up for themselves. This personality prototype portrays a lively, assertive, and socially outgoing preschooler who is unlikely to withhold his or her thoughts and emotions from others. Consistent with the finding reported by Hart et al. (1997), our undercontrolled factor did not include the elements of interpersonal exploitation reported by Robins et al. (1996) for their undercontrolled type. Robins et al. (1998) subsequently distinguished between two subtypes: the antisocial and the impulsive undercontroller. The resilient undercontroller, largely free of antisocial and interpersonally antagonistic tendencies, appears similar to the impulsive type. The relatively high level of resiliency may explain this absence of undercontrol-related problems. Finally, the third personality prototype, brittles, was characterized by CCQ items such as shyness, hypersensitivity, and anxiety, and not by items indicative of a calm, empathic, and self-reliant mode of relating to others. Replicability of the Personality Prototypes Across Studies Next, we compared our three preschool personality prototypes with the typologies identified by Asendorpf and van Aken (1999), Hart et al. (1997), and Robins et al. (1996).3 To obtain a reliable estimate and to reduce the influence of sample fluctuations, we averaged, separately for each type, the person factor scores obtained in
these three other studies and then correlated the resulting three mean person factor scores with our three personality prototype scores. The resulting convergence correlations were .81 for the first, .71 for the second, and .53 for the third personality prototype. The declining size of the convergence correlations is consistent with previous studies; the highest correlation has always been obtained for the first factor and the lowest correlation has always been obtained for the third factor. This decline in size may result from the lower reliability and poorer definition that can be expected to characterize
later extracted factors (e.g., Hart et al., 1997). In sum, good convergence was found with other studies, although our first factor was slightly more overcontrolled, our second factor slightly more resilient, and our third factor slightly less overcontrolled than the factors reported by Asendorpf and van Aken (1999), Hart et al. (1997), and Robins et al. (1996).
SUMMARY
Consistent with previous studies that used inverse factor analysis to derive personality typologies, this study also identified three preschool personality prototypes. Strong to moderate replicability was obtained. This study provides further evidence for the existence of three reasonably similar personality (proto)types, the nature of which transcends age and national origin.4 The types identified by Robins et al. (1996) in 13-yearsolds may be present in some form already in preschool. Part 2: The Construct Validity of the Preschool Personality Prototypes: Testing Directional Hypotheses Using Laboratory Test Data
The three personality prototypes were characterized by various configurations of ego resiliency and ego undercontrol. This configurality may reduce the correlations between the prototypes and procedures developed to represent relatively pure indices of each ego construct. For example, although children typifying the resilient undercontrolled personality prototype should, theoretically, show low impulse control on ego undercontrol tasks, the degree of undercontrolled behavior might be tempered by their level of ego resiliency. Nonetheless, each personality prototype is largely defined by one ego construct and predictions were therefore made on the basis of this dominant ego component. Given that ego resiliency and IQ overlap conceptually (Block&Kremen, 1996), predictions also were advanced for IQ. Thus, children typifying the overcontrolled resilient personality prototype were expected to perform well on ego resiliency and IQ measures but show no (or only moderate) relations to ego undercontrol procedures, children typifying the resilient undercontrolled personality prototype were expected to score high on ego undercontrol procedures and show no (or only moderate) relations to ego resiliency or IQ measures, and children typifying the brittle personality prototype were expected to perform poorly on ego resiliency and IQ measures. Table 3 presents the correlations between the participants’ loadings on the three personality prototypes and the measures of ego resiliency, ego undercontrol, and IQ.
THE OVERCONTROLLED RESILIENT
PERSONALITY PROTOTYPE
As anticipated, this personality prototype correlated positively (p < .05, or beyond, two-tailed) with the motor inhibition, Lowenfeld mosaics, and IQ procedures. Although the correlation with egocentrism was in the anticipated direction, it was not significant. The negative correlation with the inability to delay gratification task was not expected, but the direction of this correlation was consistent
THE RESILIENT UNDERCONTROLLED
PERSONALITY PROTOTYPE
As anticipated, this personality prototype correlated positively (p < .05, or beyond, two-tailed) with the actometer task but the correlation with inability to delay gratification, although positive, was not significant. As expected, this prototype was unrelated to measures of ego resiliency and intelligence.
THE BRITTLE PERSONALITY PROTOTYPE
As anticipated, this personality prototype correlated negatively (p < .05, or beyond, two-tailed) with the egocentrism, Lowenfeld mosaics, and IQ procedures.
Against predictions, it correlated negatively with the actometer task.
SUMMARY
Relations between the three personality prototypes— based on observer data—and the laboratory procedures indexing resiliency, undercontrol, and IQ showed substantial but not perfect convergence. A directional prediction (r = – or +) was developed for 10 of the 18 possible relations. Of these 10 relations, 7 were statistically significant (p < .05, or beyond) and in the predicted direction. For 8 of the 18 relations, a nonsignificant correlation was
expected. Of these 8 relations, only 2 achieved statistical significance: The overcontrolled resilient prototype correlated negatively with the inability to delay gratification task and the brittle prototype correlated negatively with the actometer task.
Part 3: Developmental Implications of Preschool Personality Prototypes for Adjustment in Adolescence The participants typifying the overcontrolled resilient
personality prototype were anticipated to remain intelligent and overcontrolling in adolescence. They also were expected to be less resilient in adolescence
than in preschool, mainly because the overcontrolling component of this personality prototype was hypothesized to reduce their ability to cope with adolescent challenges (Block&Gjerde, 1986). Undercontrol of impulse has been found to remain relatively longitudinally consistent (Block & Gjerde, 1986; Caspi et al., 1995). The participants typifying the resilient undercontrolled prototype were therefore expected to remain undercontrolled in adolescence. Finally, participants typifying the brittle personality prototype were expected to remain maladjusted and relatively unintelligent; their lack of resiliency was anticipated to minimize their ability to cope with the challenges that characterize adolescence. To determine the relations between preschool personality prototypes and adolescent adjustment, correlations were computed between the personality prototype loadings and (a) the 100 age-14 CAQ items, (b) the personality constructs, (c) the age-11 and age-18 IQ, and (d) the age-14 drug use measure. The number of sex differences in the correlation patterns did not exceed the number to be expected by chance. We therefore only report results for the total sample.
The overcontrolled resilient prototype was significantly associated (p < .05 or beyond) with 33 CAQ items, the resilient undercontrolled prototype with 34 CAQ items, and the brittle prototype with 6 CAQ items.5 The significant items were grouped using principal component analysis. The resulting components were not introduced as replicable factors but only to aid the reader in assimilating the findings; the aim is data reduction not “dimensional discovery” (Lykken, 1971). In principal component analysis, the components are merely linear combinations of the observed variance; thus, underlying hypothetical factors should not be postulated.We therefore refer to the CAQ item correlates of each prototype grouped by means of principal component analysis as “themes,” not as “factors.” The CAQ-based personality themes rendered a psychologically coherent portrait of the first two personality prototypes. Table 4 presents the
relations between the three personality prototypes and the personality constructs, IQ, and drug use.
LONGITUDINAL PERSONALITY CORRELATES OF THE OVERCONTROLLED RESILIENT PROTOTYPE
Three CAQ themes, reclusiveness (16 items, α = .95), introspective acquiescence (7 items, α = .92), and dependable resourcefulness (10 items, α = .85), summarized the CAQ correlates of the overcontrolled resilient prototype. Reclusiveness was positively defined by such observer assessments as “is emotionally bland,” “communicates through action,” and “tends to ruminate” and negatively defined by such CAQ items as “is a talkative individual,” “initiates humor,” and “is facially, gesturally expressive.” Introspective acquiescence was positively defined by such observer assessments as “overcontrols needs, impulses,” “introspective, concerned with self as an object,” and “has readiness to feel guilt” and negatively defined by such CAQ items as “is assertive” and
“tends to proffer advice.” Dependable resourcefulness was positively defined by such observer assessments as “has high degree of intellectual capacity,” “favors conservative values,” and “is calm, relaxed in manner” and negatively defined by such items as “unable to delay gratification,” “tries to stretch limits,” and “is self-indulgent.” The more the participants fit the overcontrolled resilient personality prototype as preschoolers, the more likely they were, as adolescents, to be shy and conscientious and less likely they were to be undercontrolling, active, and extraverted. The correlations with age-11 and age-18 IQ were positive, whereas the correlation with drug use was negative and marginally significant. Summary of the overcontrolled resilient personality prototype. The adolescent attributes associated with this prototype portrayed a relatively well-adapted yet emotionally bland and socially reclusive youth characterized by little overt psychological conflict. These adolescents were solitary, conscientious, and cautious; they were not vivacious or socially at ease. They were nonetheless resourceful, intelligent, and receptive to their own feelings. There was some evidence that their bland social stimulus value concealed access to a differentiated inner life. Why the absence of ego resiliency in the adolescents who typified this preschool personality prototype?
Overcontrolled resilience may have served the participants well in the relatively predictable niches of early childhood. Yet, the overcontrolling component may have reduced their ability to cope resourcefully with the challenges and emotional turbulence during the transition from childhood into adolescence. Second, due to their behavioral inhibition in preschool, the individuals typifying this prototype may have gradually withdrawn from social participation. Their reclusiveness may have limited their opportunity to learn new coping skills relevant to the adolescent adaptation, leading to a decline in their level of resiliency. This pattern exemplifies proactive interaction, whereby individuals select or create their own environments (Caspi&Bem, 1990). Third, the
nursery school teachers also may have overestimated the level of resiliency of these children because of their dependability, helpfulness, and compliance. This pattern may be less advantageous later in life (see also Asendorpf & van Aken, 1999; Gjerde, 1995). Although we are not able to decide among these explanations, undercontrol showed, as anticipated, greater longitudinal consistency than resiliency.
LONGITUDINAL PERSONALITY CORRELATES OF THE RESILIENT UNDERCONTROLLED
PERSONALITY PROTOTYPE
The CAQ correlates of the resilient undercontrolled personality prototype were summarized by three CAQ themes: sociability (15 items, α = .95), assertive independence (10 items, α = .92), and inability to delay gratification (9 items, α = .91). Sociability was positively defined by such observer assessments as “skilled in social techniques,” “initiates humor,” and “is self-dramatizing, histrionic” and negatively defined by such CAQ items as “does not vary roles,” “is emotionally bland,” and “uncomfortable with uncertainty.” Assertive independence was positively defined by such observer assessments as “is assertive,” “tends to proffer advice,” and “values independence, autonomy” and negatively defined by such CAQ items as “submissive, accepts domination,” “has a readiness to feel guilt,” and “vulnerable to threat.” Inability to delay gratification was positively defined by such observer assessments as “unable to delay gratification,” “tries to stretch limits,” and “tends to be rebellious, nonconforming” and negatively defined by such CAQ items as “overcontrols needs, impulses,” “favors conservative values,” and “dependable, responsible person.” The more the participants fit the resilient undercontrolled personality prototype as preschoolers, the more likely they were, as adolescents, to be undercontrolled, active, and extraverted and less likely they were to be shy and conscientious. The correlations with age-11 and age-18 IQ were insignificant, whereas the correlation with drug use was moderately positive.
Summary of the resilient undercontrolled personality prototype. The adolescent attributes associated with this prototype indicated continuation of low impulse control. As adolescents, individuals typifying this personality prototype were sociable, assertive, extraverted, and somewhat lacking in conscientiousness. This pattern resembles the lack of control type (Caspi et al., 1995). Ego undercontrol also may have temperamental underpinnings (e.g., Rothbart & Ahadi, 1994). Although undercontrol is sometimes associated with vulnerability (e.g., Block’s, 1971, description of unsettled and vulnerable undercontrollers), the resilient undercontrolled prototype emphasized the positive aspects of this attribute, including energy and social skills. Preschool children typifying this factor did not turn into antisocial adolescents but rather into youth who functioned well interpersonally; they did not show the behavior problems that characterized Robins et al.’s (1996) undercontrolled type. Although the correlation with drug use was positive, it was low. This difference may result from the large representation of high-risk boys in the sample used by Robins et al. (1996). High self-esteem and relative absence of antisocial tendencies also characterized Hart et al.’s (1997) undercontrolled children. In adolescence, as in preschool, the individuals typifying
this personality prototype illustrated the impulsive rather than the antisocial component of undercontrol (see Robins et al., 1998, for a further description of this distinction). In sum, the individuals typifying this prototype may have adjusted to the world in a manner not pulling them toward change.
LONGITUDINAL PERSONALITY CORRELATES OF PRESCHOOL BRITTLE PERSONALITY PROTOTYPE
The very few significant age-14 CAQ correlates associated with this prototype may be due to chance and were not interpreted. The correlations with IQ remained negative. Summary of the brittle personality prototype. This prototype predicted few developmental outcomes with the exception of relatively low intelligence. Hart et al. (1997) also reported relatively weak results for their third type. As noted above, factors extracted late may be conceptually less well defined and less reliable than factors extracted early. In addition, relatively few individuals received high loadings on this personality prototype and the rank order of the many low loadings may not be reliable. Correlations involving this factor might, therefore, for psychometric reasons, not be strong. Internalizing problems—akin to those characterizing brittle and overcontrolled individuals—are also likely to fluctuate over time (Ollendick & King, 1994). Both psychometric and theoretical issues may therefore have contributed to the weak developmental outcomes for this prototype.
DISCUSSION
This study examined four issues: (a) Could psychologically coherent personality prototypes be discerned as early as in preschool (i.e., the issue of coherence)? (b) Did our preschool personality prototypes converge with personality prototypes identified in samples of slightly older participants (i.e., the issue of replicability)? (c) Did individuals typifying each personality prototype show theoretically meaningful relations to performance on experimental tasks (i.e., the issue of construct validity)? and (d) Did individuals typifying each personality prototype show different developmental outcomes (i.e., the issue of developmental implications)? Three psychologically coherent personality prototypes were identified. Strong to moderate convergence
with typologies from other studies was observed. Children typifying each personality prototype showed theoretically meaningful, although less than perfect, relations to experimental procedures of ego resiliency, ego undercontrol, and IQ. Preschool children typifying the overcontrolled resilient factor were, as adolescents, shy and inactive, but also dependable, conscientious, and intelligent. In contrast, preschool children typifying the resilient undercontrolled personality prototype were, as adolescents, active, sociable, nonconforming, and extraverted. As expected, impulse control showed greater longitudinal consistency than resiliency. This study adds to a growing literature indicating that typologies developed as early as preschool anticipate adolescent adjustment (e.g., Asendorpf & van Aken, 1999; Caspi & Silva, 1995). We recognize that the size of the correlations between the preschool personality prototypes and the personality outcomes at age 14 were moderate. Caspi and his colleagues in their typological study of 3-year-old children also reported relatively small effect sizes over time (Caspi, 2000). However, in evaluating these longitudinal results, two points should be recognized. First, personality outcomes in adolescence are likely to be influenced by factors other than the constellation of personality attributes in preschool. Ahadi and Diener (1989) showed why effect sizes, by mathematical necessity, often have to be small when the outcome variable is multidetermined. Second, “because the effects of personality differences accumulate over a lifetime, a focus on a single point in time will result in an underestimate of the extent of continuity in behavioral development” (Caspi, 2000, p. 168). Is There a Replicable Typology of Persons? Although this study identified personality prototypes comparable to those found in other studies, personality prototypes must not be prematurely reified as sampleindependent personality descriptors or as authentic substantive entities. The Q-factor approach is sensitive to the sampling of persons and the results may therefore describe samples as much as individuals. Even though similar verbal labels are used in different studies, the exact nature of corresponding types nonetheless varied. For example, although the undercontrol label has been used in several studies to identify a specific type (e.g., Hart et al., 1997; Robins et al., 1996), its most defining personality attributes still differed across studies despite moderately high convergence correlations. This lack of complete content overlap occurred even for the first and most replicable factor. Although our first factor correlated .75 with Robins et al.’s (1996) first factor, only 5 CCQ items were included among the 10 most and 10 least characteristic items that defined this factor in both studies. Even high factor convergence correlations may conceal tangible differences.
The strongest evidence for the existence of these three personality prototypes comes from studies that have used the California Child Q-sort. The CCQ was developed to provide extensive coverage of ego resiliency and ego control. It is therefore encouraging that three personality prototypes were closely recaptured by Harrington, Chin, Rickey, and Mohr (1999), who used Gough and Heilbrun’s Adjective Check List (Gough & Heilbrun, 1965). Thus, the three personality prototypes (resiliency, undercontrol, and overcontrol) do not seem uniquely dependent on the CCQ (see also Asendorpf & van Aken, 1999, p. 830). Nonetheless, the argument that these three types constitute the basis for any typology of persons may prematurely foreclose search for different typologies.
Similar to most other person-centered studies (e.g., Asendorpf & van Aken, 1999; Hart et al., 1997; York & John, 1992), this study examined typologies identified in a single point in time: at the beginning of the study. This approach is useful when the goal, as in this research, is to predict outcomes based on early existing attributes. A typology developed at a single point in time—either at the beginning or at the end of the study—has limitations.
First, the nature of the typology itself may depend too strongly on the time period during which it was developed—early childhood typologies may not be identical to adult typologies. Second, the “single-point” approach does not build life trajectories into the very definition of the typologies themselves. Gjerde, Chang, and Kremen (1998) developed typologies simultaneously identified at three separate ages, each a decade apart. These developmental typologies, simultaneously based on data from different developmental periods, provide more differentiated portrayals of life trajectories and illustrate how the person-centered approach may bridge personality and developmental research. Methodological Caveats Consistent with the majority of person-centered studies,
inverse factor analysis was employed to generate the personality prototypes. In contrast to R factor analysis, inverse factor analysis requires more variables than participants (Bailey, 1994). This assumption is seldom satisfied. Given the high number of participants relative to variables in most person-centered studies, the high replicability of the three personality prototypes across studies is promising. These findings notwithstanding, students of person typologies might benefit from multiple classification methods to avoid placing “all research eggs . . . in the same methodological basket” (Nunnally, 1978, p. 625). Other methods have shown promise (e.g., Fraley & Waller, 1998). This study identified personality prototypes at a single time period. It can therefore not estimate the continuity of the personality prototypes or the stability of group membership across time. Type membership may change over time (e.g., Gjerde et al., 1998). The emergence of personality types is also a crucial issue for future research. The identification of the conditions under which group membership changes (or remains stable) may help us understand the developmental unfolding of different types (e.g., Hart et al., 1997).
Coda
Developmental and personality research would seem to benefit from a greater focus on (a) the person as a basic analytical unit and (b) the diverse developmental pathways characterizing different homogeneous groups of individuals. However, the person-centered concept carries different meanings to different researchers. Consistent with previous studies (e.g., Asendorpf & van Aken, 1999; Hart et al., 1997; Robins et al., 1996; York & John, 1992), this study used inverse factor analysis to identify coherent and replicable personality prototypes. But once these prototypes were identified, they were—as in similar studies—used in a variable-centered way in that person loadings were correlated with other variables. Hart et al. (1997) noted that once personality types have been identified, intellectual inquiry necessarily must return to a variable-centered approach. For example, is “the undercontrolled child’s aggressiveness primarily the consequence of a need to defend extremely high self-esteem?” (p. 204). This study may therefore be viewed as combining both person-centered and variable-centered analytical strategies. In general, the relationship between variable-centered (or nomothetic) laws, which holds for groups, and ideographic laws, which holds for specific individuals, may be both more complex and less contradictory than generally believed. Meehl suggested in 1954 that “behavioral laws are (a) nomothetic in their form for a given group, (b) ideographic in their parameters, and (c) strongly ideographic in their end terms, which refers to response properties of the organism” (Wiggins, 1973, pp. 148- 149). Allport (1937), as well, did not argue that general laws precluded uniqueness. Wiggins (1973, Chap. 4) provides an excellent discussion of these issues. Some psychologists may not consider this study as truly person centered. We grouped participants into three homogenous personality prototypes. Future studies, however, might do well to include case examples of each personality prototype. Two individuals who both received high loadings on the same personality prototype may still have different life stories to tell, construct their lives differently, or hold different personal goals. Future research might benefit from integrating personality typologies with personal concerns and narrations of the self (i.e., Level II and Level III of personality, as described by McAdams, 1995). Furthermore, identification of personality prototypes in combination with in-depth case studies representative of each prototype might bridge the personological tradition—with its focus on individuals in their life context—with the approach taken in this study. Greater reliance on narrative methodologies appears to be a particularly promising avenue to pursue to achieve this aim.