Parental Descriptions of Child Personality

Developmental Antecedents of the Big Five?

Gedolph A. Kohnstamm, Charles F. Halverson, Jr., Ivan Mervielde, Valerie L. Havill, Charles F. Halverson, Gedolph A. Kohnstamm, Charles F. Halverson, Jr., Ivan Mervielde, Valerie L. Havill, Charles F. Halverson

• Publisher: Psychology Press
• Year: 1998
• ISBN:
  9781135690007
• Edition: 1
• Topic: Psychology
• Subtopic: History & Theory in Psychology
• Index: Psychology

This book reports the first attempt in the child development literature to examine the structure of early personality based on parents' free-descriptions of their children. It is an important piece of research because of its cross-national focus on personality development. The authors present a data set that reveals considerable consistency in the parental descriptions of child personality in both western and nonwestern countries. This consistency supports the cultural universality of the "Big Five" personality factors. The authors' findings lay the foundation for an examination of how these major dimensions of childhood personality structure evolve into adult personality structure.

1
Analyzing Parental Free Descriptions of Child Personality

Geldolph A.Kohnstamm

Leiden University

Charles F.Halverson, Jr.

University of Georgia

Ivan Mervielde

University of Ghent

Valerie L.Havill

University of Georgia

Since Thomas and Chess popularized the concept of temperament in childhood (e.g., 1977), many studies devoted to temperamental differences among children have been published, and the nine-dimensional structure devised by Thomas, Chess, and their collaborators has attained textbook status. Applied initially to the study of infants, but later to older children as well, this approach categorizes temperamental differences based on clinical experience into nine more or less independent traits: Activity Level, Rhythmicity, Approach-Withdrawal, Adaptability, Threshold of Responsiveness, Intensity of Reaction, Quality of Mood, Distractibility, and Persistence. Beyond these first-order constructs, second-order constructs of three clusters—easy, difficult, and slow to warm up—have been proposed based on factor analyses of the nine scales.

Because of their clinical usefulness, the temperament scales developed by Thomas and Chess and later by Carey and others (e.g., Carey & McDevitt, 1978) became well-known instruments for assessing temperamental differences in infancy and childhood. We have begun with this brief history to emphasize that temperament measures currently in use, regardless of format, are topdown, theoretically derived instruments, with items reflecting the concerns of the child-care specialists. Even when some form of factor analysis has been used to summarize dimensions statistically—as opposed to clinically—the final result has still depended on item selection done in constructing the instruments. In the case of the scales by Thomas and Chess and by Carey (e.g., McDevitt & Carey, 1978), other authors have shown that the original nine dimensions have not been recovered from factoring at the item level. Instead, from five to seven factors not closely resembling the original nine have emerged (see Martin, Wisenbaker, & Huttunen, 1994). We have emphasized here how the theoretical and clinical concerns of the Thomas and Chess group have shaped the temperament field. We have no quarrel with the usefulness of the nine-dimensional structure for many clinical assessments or research programs, but the dimensions might not be comprehensive, coherent statistically, or robust across ages and cultures. There are other ways to develop items, as we next propose, and there is a pressing need to assess the cross-cultural generality of constructs developed largely in one language or culture. Anthropologists and cross-cultural researchers have begun to demand justification for such cross-cultural applications of psychological instruments (e.g., Malpass & Poortinga, 1986; Shwalb, Shwalb, & Shoji, 1994).

Although it might seem that cultural homogeneity is on the increase at the same time that cultural uniqueness is on the decrease, a world where cultural differences have vanished in one big melting pot has not yet arrived. Thus, the appropriateness of translation and application of psychological instruments across cultures should be questioned more frequently than is presently done. Thus far, cross-cultural studies in the field of temperament have consisted mostly of comparisons of means and variances on scales originating in England (e.g., Eysenck Personality Questionnaire [EPQ], Eysenck & Eysenck, 1975) or the United States and translated into other languages. Although this work has produced many interesting results, we decided to follow a different approach.

FREE DESCRIPTIONS OF PERSONALITY

We were originally motivated by John (1990a), who called for more studies using free descriptions of personality to test for the possibility that the reliance on top-down, theorist-imposed descriptors was “too parochial”:

…given that the Big Five were intended to represent the major dimensions of natural-language personality descriptions, another option is to investigate the characteristics people use in free descriptions of themselves and others. Would the Big Five be replicated if the set of descriptors factored was based on the content of subjects’ free descriptions, rather than on those sets of terms selected by the taxonomers themselves? (p. 92)

To explore this question, John and Chaplin (John, 1990a, 1990b) asked more than 300 U.S. college students to describe their own personalities and to generate terms for both their desirable and their undesirable characteristics. This first phase of collecting, categorizing, and counting descriptors was then followed by a second phase in which the 60 most frequently used descriptors were put in a questionnaire and given to a new sample of subjects. Factor analyses of self-ratings yielded five factors that closely resembled the conceptual definitions of the Five-Factor Model (FFM), a finding supporting the hypothesis that these five factors are indeed the most salient dimensions of personality for U.S. college students.

The FFM free-description methodology has been used cross-culturally as well. For example, Church, Katigbak, and Castaneda (1988) conducted in-depth, open-ended interviews with 41 Filipino bilingual (English and Tagalog) college students. The students provided general descriptions of healthy and unhealthy Filipinos in several broad areas of functioning (e.g., attitudes and feelings toward others; actions with others; attitudes and feelings toward humans in general; attitudes, feelings, or thoughts about themselves; goals or values; and mood). Responses were recorded verbatim and transcribed.

The 1,516 nonredundant descriptors obtained were inductively reduced to 54 semantic categories. The authors could allocate almost all of these 54 categories to one of the five dimensions of the FFM. There were, however, additional dimensions not easily summarized by the FFM. For example, a dimension relating to nationalism and societal awareness emerged. Although these descriptors might fit in Category III (Conscientiousness) in the FFM, the authors suggested that their saliency in the personality descriptions was in response to the emphasis on social and political awareness in the Philippines during the time of the research.

In these studies, spontaneously mentioned personality characteristics were valued because it is assumed that people frequently mention those characteristics that they think are most important or basic. Further, it is assumed that aggregating these spontaneously mentioned descriptors over groups of individuals yields a set of perceptions specific to the culture of the informants.

Coding Free Descriptions

The collection and categorization of free descriptions of personality are the first steps in our research program. When many people have been interviewed, the large collection of personality descriptors obtained must be ordered. This is done by judges using a well-tested categorization system with good interjudge reliability.

When the system of categories has been applied, we can assess category frequency. Then, assuming that frequency of category use indicates the degree of saliency of each personality category in a particular culture, we choose exemplars from the high-frequency categories to prepare for the second phase. In this phase, the representative selection of descriptors is put into a questionnaire format that is then given to new samples of people from the particular culture involved. They may be asked to rate themselves or others on the items selected. Finally, factor analyses of the items are used to summarize the underlying dimensions in the set of characteristics.

Free Descriptions of Child Personality

In the late 1980s, in his own search for the major dimensions in parental perceptions of temperament and personality in children, Kohnstamm and his students did some pilot work using parental free descriptions. These pilot studies served to develop and refine the methodology of eliciting and coding parental free descriptions. The studies demonstrated the sensitivity of the free-response format to detecting both social class and informant differences (e.g., mother versus father) in the Dutch language and culture, and they also revealed the potential of the FFM for categorizing parental descriptors. Until 1990, only one investigator had explored the structure of perceived personality in late childhood and adolescence from a FFM perspective (Digman, 1963, 1990; Digman & Inouye, 1986).

During a sabbatical year together at the Netherlands Institute for Advanced Study (NIAS) in 1990–1991, Kohnstamm and Halverson formulated the idea for the project described in this volume. An international group of researchers was formed with the goal of collecting free descriptions of children of different ages, in different languages and cultures. We assumed that our method of collecting free descriptions would tell us whether we would get the same or different personality dimensions over different languages and cultures. To quote from Church and Katigbak (1989): “By starting with a taxonomy of personality concepts generated independently in each culture, culture-relevant dimensions are allowed to emerge independently, providing a more convincing test of universality when comparable dimensions emerge” (p. 870).

Although the societies represented in the present study are all modern, the families involved live in differing social and economic circumstances, with large differences in household income and in future prospects for the children described. With the samples of differing cultures and circumstances, we can assess whether the parents in these societies generate temperament and personality descriptors with differing emphases, frequencies, and evaluations. When questionnaire items are distilled from the variety of terms, we can assess whether different dimensions emerge from parents’ perceptions of their children’s personalities in these different cultures.

METHOD

In all participating countries, personality descriptions were collected from parents of children between 2 and 13 years of age. In some samples, parents were simply asked to tell us about their children. In other samples, after an introduction in which the word “personality” was mentioned, the parents were asked, “Can you tell me what you think is characteristic of your child?” All interviews were audiotaped and subsequently transcribed verbatim. The coders used elaborate coding manuals that included instructions about units of analysis, division of phrases, dealing with repetitions, and synonymy.

For our purposes, a unit of analysis was defined as an adjective, verb, noun, or phrase referring to a description of behavior, personality characteristic, or ability. Phrases referring to situational causes of behavior or to physical attributes were not coded. Because a unit of analysis could be a phrase, it was sometimes helpful to split phrases into simple, easily codable parts. Adjacent words or phrases could be divided and coded separately as two individual units if the meaning of each part was understood when considered independently. If a coder judged that meaning or context was lost by splitting the phrase, the unit was coded as one single description. For example, the phrase “She likes to play outdoors with neighbor kids” can be separated into two distinct parts: “She likes to play outdoors” and “Plays with neighbor kids.” The first phrase would be coded as referring to physical activity level, and the second phrase would be coded as indicating extraversion or sociability. The phrase “She’s so quick; her head works very, very fast” would be coded as a single unit because breaking the description into two parts could conceivably lead the coder to misinterpret “She’s so quick” as referring to physical activity instead of cognitive proficiency.

In free-language interviews, respondents often elaborated on a single characteristic by mentioning concrete, situation-specific behaviors to illustrate the personality characteristic. In such cases, the elaborative phrase or phrases were taken with the descriptive word or phrase and were coded as one unit of analysis. Respondents might also mention a descriptive characteristic in the past tense and contrast this with a similar descriptive characteristic in the present tense to illustrate the way a child is now with respect to a younger age. In this case, the past-tense phrase was not coded separately, but was included in one unit with the presenttense phrase. The part of the phrase in the present tense was the subject of analysis; the past-tense word or phrase, however, might have helped the coder to assess the meaning or importance of the unit as a whole.

Words and phrases that were not coded as descriptive phrases included those referring to a person other than the target child or to children in general. These were considered nonrelevant phrases. Phrases about peripheral information were also excluded; these included information connected to the main issue, but so remote as to have no immediate relevance to the target child (e.g., “Her parents are friends of mine” or “You have a lot of temper tantrums and things with all kids, you know; that is not specific to Susie, but that is something that would bother me about her”). If there was reasonable doubt as to whether a respondent was referring to the target child directly, then the word or phrase was not coded.

When words or phrases were repeated verbatim or if phrases expressing the same literal meaning were used more than once in a single interview, these units were recorded and coded as repetitions, but were not included in frequency analyses more than once.

The Categorization System

To categorize the expressions generated by the parents, a coding system was developed. Although the system of categories was inspired by the FFM framework, with several subcategories in each of the five dimensions, an additional eight categories were added. Each major category was designated by a Roman numeral. The first five were numbered according to conventions in the FFM literature. The subcategories or facets were inductively derived. Responses are coded as Positive, Neutral, or Negative as well. For example: “Enthusiastic” is coded as IA+; “Tends to shut herself off” is coded IA−. Table 1.1 shows the total system. No examples are given for responses coded as Neutral. Decisions were made on which utterances were to be used as units for coding and which utterances could be discarded. Detailed instructions were developed for this step in applying the categorization scheme.

Rationale for Categories Included

The origin of the first five main categories has already been explained. The subcategories within these five are our own inventions, based on clusters of high-loading items, or “facets,” as published in several FFM studies. For the location of some of the subcategories (e.g., Manageable for Parents and Teachers—Category II), we had no empirical basis: No adjectives for Manageability were included in FFM adjective studies as they did not deal with children. In the instrumentation phase of this project, items dealing with Manageability might not cluster with a higher order factor recognizable as the FFM’s Agreeableness. We emphasize here that we do not necessarily expect to find a neat FFM structure once new samples of

TABLE 1.1 Categories for Coding Descriptors From “Free” Personality Descriptions and Examples of Descriptors

As for the rationale for the coding in our pilot studies for the categories additional to the FFM (i.e., Categories VI through XIV), we coded the Independence (VI) category separately from the Big Five for two reasons. First, when parents described their children as being independent (or as being too dependent), they might mean something different from being simply high or low on Extraversion, Conscientiousness, Emotional Stability or Instability, or Openness to Experience. Second, in the FFM literature about adults, a factor labeled as Autonomy or Independence has been repeatedly seen as having independent status. So, on rational grounds, John (1990a, 1990b) made a separate category for Independence when categorizing personality descriptors generated by his students. Also, Costa and McCrae (1988), on empirical grounds, saw sufficient indications for a separate factor of Autonomy in the Personality Research Form (PRF). Whether being independent and autonomous in childhood will attain independent factorial status remains to be determined in the following phases of our project.

Mature for Age (VII) is a category specific to children. Very few adjectives of this kind have ever been included in FFM adjective studies using self- or other ratings of adults, and it is therefore impossible to tell if and where Mature for Age would fit in the factor analytically derived model. The category is included for coding comprehensiveness and possible links to other categories.

It is questionable whether a person’s being often or never ill or having a disability is a personality characteristic in the strict sense. We have included Illness, Handicaps, and Health (VIII) in our system because parents of children who are ill or disabled often mentioned this fact first when beginning the interview. These parents considered their children’s condition as fundamental background information for understanding the youngsters’ other characteristics.

Rhythmicity (IX) is included as a coding category because it is one of the nine dimensions of the Thomas-Chess model. In the Thomas-Chessderived DOTS-R questionnaire (Revised Dimensions of Temperament Survey; Windle & Lerner, 1986), Rhythmicity is even operationalized in three separate scales (for eating, sleeping, and daily habits). Angleitner and Ostendorf (1994) demonstrated that when students rate themselves on many different personality questionnaires, including the DOTS-R, Rhythmicity obtains independent status as a sixth factor outside the FFM domain. We found, however, that only a very small proportion of the descriptors had to do with Rhythmicity. Parents from other cultures might possibly generate more descriptors indicating aspects of Rhythmicity than we have found so far.

Gender-Appropriate Behavior and Physical Attractiveness (X) are concepts not usually included in the measurement of temperament and personality, although they are important personal characteristics that could cluster with other major personality traits in childhood. Usually parents, at least in Western cultures, do not mention physical attractiveness (or lack thereof) in a conversation with a stranger. Nevertheless, we coded descriptors for physical attractiveness because of its importance for both children and adults. For example, Lanning (1994) found an independent attractiveness factor, the first one after the Big Five, when factoring a sample of 940 California Adult Q-Set (CAQ) ratings of students in California.

We also included a category for descriptors indicating how well children are doing in school, for example, whether their marks are good, average...