The Nature Etiology and Consequences of Individual Differences in Temperament

Theodore D. Wachs


Over the past 10 years the increasing importance of temperament as both a critical developmental outcome and as a moderator and predictor of other developmental outcomes is mirrored by the increasing number of books and major review chapters devoted to this domain (e.g., Guerin, Gottfried, Oliver, & Thomas, 2003; Halverson, Kohnstamm & Martin, 1994; Molfese & Molfese, 2000; Rothbart & Bates, 1998; Wachs & Kohnstamm, 2002). Given this wealth of information, a very obvious question is: Does the world really need another chapter on temperament? Answering this question in the affirmative is based on what appears to be a shift in our understanding of the nature and consequences of individual differences in temperament. Historically, with some notable exceptions, conceptualization of and research on the nature and consequences of individual differences in temperament have focused on temperament as a main effect predictor and outcome. A fundamental thesis of this chapter is that, by focusing just on temperament as a single main effect predictor or outcome, we severely limit our ability to deal with critical issues such as the delineation of the domains of temperament, discordance between parent reports of their child's temperament, the modest stability of temperament traits over time, the etiology of individual differences in temperament, and the consequences of individual differences in temperament. Rather than viewing temperament in isolation, I will argue it is essential that we view temperament as one part of a system of linked multiple influences and outcomes. Other parts of this system include:

1. Other developmental domains like cognition and motivation;

2. Nontemperament child characteristics like age, gender, and biomedical or nutritional status;

3. Changes in the structure and functioning of the central nervous system or gene action patterns;

4. Different levels of the individual's context including:

a. proximal environmental characteristics such as parental sensitivity, involvement, and responsiv-ity, as well as environmental "chaos" in the home;

b. quality of parental marital relationship;

c. distal environmental characteristics such as cultural values and beliefs about the desirability of various child characteristics, or the availability of environmental "niches" that are open to the individual.

As will be discussed, by viewing temperament as part of a system rather than in isolation, we are far better able to understand the nature and changes in different dimensions of temperament, how best to measure temperament, and the role temperament plays in various aspects of individual development.

The seeds for both the mainstream approach to temperament in isolation and the alternative view of temperament as part of a system were sown in the pioneering work of Thomas and Chess (Thomas & Chess, 1977; Thomas, Chess, & Birch, 1968). A major rationale underlying the New York Longitudinal Study of Thomas and Chess was to document how the prevailing view of the etiology of behavior disorders—inadequate or inappropriate parenting—was insufficient given the potential importance of biologically based individual differences in children's behavioral styles (temperament). By contrasting temperament as etiology versus parenting as etiology, Thomas and Chess laid the groundwork for later researchers who focused on temperament as an isolated influence. One result has been the increasing number of studies which have focused on two variable questions, with temperament serving as either the predictor or the outcome. Questions addressed by these types of two variable studies include whether specific temperament dimensions are heritable, whether individual differences in attachment reflect individual differences in temperament, whether differences in child temperament produce differences in parent behavior, whether parent reports of children's temperament primarily reflect parent characteristics, the link between early temperament and later personality and how well children with certain temperaments perform in school or test situations (for a review of evidence on these questions see Molfese & Molfese, 2000 or Rothbart & Bates, 1998). For example, there have been multiple studies using point-to-point correlations which have focused on the stability of specific dimensions of temperament or on the degree of linkage between temperament and Big Five personality dimensions. In much of this research, little consideration has been given to nontemperament influences which could act to moderate or mediate observed correlations (Wachs, 1994).

While Thomas and Chess may have sparked our focus on temperament in isolation, in their research and conceptualization they have always been careful to emphasize that the developmental consequences of individual characteristics like temperament cannot be understood without detailed consideration of the context within which an individual with specific temperament characteristics functions—their well-known concept of "goodness of fit." By focusing on the match between temperament and context as the major contributor to individual differences in adjustment Thomas and Chess also laid the groundwork for an alternative viewpoint; namely, that we cannot understand the impact of temperament on behavior without also understanding the nature of the context within which individuals with different temperaments function. This alternative viewpoint is best reflected in studies that have looked at the joint contributions of temperament and contextual characteristics to a variety of developmental outcomes including behavior disorders (Bates, 2001; Eisenberg et al., 2001), infant cognitive performance (Halpern et al., 2001), attachment (Kochanska & Coy, 2002), and children's development of an internalized conscience (Kochanska, 1993, 1995).

There seems to be no doubt that the study of temperament in isolation was a necessary and critical step in terms of helping researchers understand the nature of temperament, the development of specific dimensions of temperament, and the contributions of temperament to development. However, I argue that the extent of our database on temperament is more than sufficient to allow us to take the next necessary step; namely, moving from the artificial world of temperament in isolation to the real world of temperament as part of a system. For purposes of the present discussion, I will use a simplified structural definition of a system; namely, a set of organized multiple elements, with each element linked to and interacting with at least some of the other elements in the set (Wachs, 2000a). As I illustrate in the following sections, viewing temperament as part of a system has two advantages. First, it allows us to deal better with certain perennial problems involving the conceptualization and measurement of temperament which have not been satisfactorily resolved by looking at temperament in isolation. Second, viewing temperament as part of a larger system allows us to develop a more comprehensive picture, both about the etiology and development of individual differences in temperament, as well as about the extent and nature of the contributions of temperament to individual variability in behavior and development. The advantages of viewing temperament as part of a larger system will be illustrated below through use of questions involving what, why, and how. Specifically, the following questions will be addressed: (a) What are the domains of temperament? (b) Why are levels of interparent agreement on child temperament characteristics and stability of temperament so modest? (c) What accounts for individual variability in the development of temperament? (d) How does temperament influence context? (e) How does individual variability in temperament translate into variability in individual developmental patterns?


At present, there is neither a single agreed on definition of temperament nor agreement on what constitutes the domains of temperament (Belsky, Hsieh, & Crnic, 1996). This problem is not unique to temperament; similar problems are also found in other areas studied by psychologists, such as intelligence. Given the social nature of science (Kuhn, 1970), the existence of multiple definitions and domains has not proven to be a major hindrance to progress as long as there is general consensus among researchers as to the major definitional features of a given construct and the domains that fit under this construct (McCall, 1986). While there are multiple definitions of temperament, for all intents and purposes the majority of researchers in the field would accept as a "working definition": "Biologically rooted individual differences in behavior tendencies that are present early in life and are relatively stable across various kinds of situations and over the course of time" (Bates, 1989, p. 4). There also appears to be agreement that the individual differences referred to in the working definition should bear some resemblance to later appearing personality traits (Halverson et al., 1994; Strelau, 1987).

While a variety of individual characteristics have been listed under the rubric of temperament, from the relatively restricted focus on emotionality by Goldsmith and Campos (1982) to the Big Nine of Thomas and Chess (1977), there does appear to be general agreement that the following list defines the major dimensions of temperament (Bates, 1989):

1. Negative emotionality (e.g., fear, anger),

2. Difficultness (e.g., high intense, easily evoked negative moods),

3. Adaptability to new situations or people (e.g., inhibition),

4. Activity level,

5. Self-regulation (e.g., soothability),

6. Reactivity (e.g., how intense a stimulus is needed to invoke a response),

7. Sociability-positive emotionality (e.g., pleasure in social interactions).

Going beyond individual dimensions there also is increasing agreement that the various temperament dimensions listed above fall into one of two overall domains: reactivity and self-regulation (Rothbart & Bates, 1998).1 Reactivity encompasses those dimensions of temperament involved in the onset, duration, and intensity of response to social and object stimulation, and can be further divided into positive and negative reactivity, with negative reactivity in turn being subdivided into anger (e.g., distress to limits) and fear (e.g., distress to novelty) (Rothbart & Derryberry, 1981; Rothbart, Der-ryberry, & Hershey, 2000). Self regulation refers to those processes that act to attenuate or accentuate the individual's reactivity to stimulation (Rothbart et al., 2000). Like reactivity, self-regulation can also be subdivided. The earliest appearing form of self-regulation is reactive control, which involves involuntary tendencies to avoid or inhibit responding to negative stimulation and to approach positive stimuli. Active control, which involves voluntary attentional regulation and the ability to inhibit ongoing behavior, appears later in development and can supersede the influence of reactive control tendencies (Eisenberg, 2002; Derryberry & Rothbart, 1997). Within this two domain framework it is the interaction between different levels of reactivity and self-regulation that acts to guide the course of temperament driven behavior (Belsky, Friedman, & Hsieh, 2001; Eisenberg et al., 2001). For example, 5-month-old infants who were high in reactivity showed different degrees of defiant behavior at 30 months, depending upon their level of self-regulation at 5 months (Stifter, Spinrad, & Braungart-Rieker, 1999).

Regardless of whether temperament is viewed at the dimension or the domain level, a continuing problem has been the difficulty in defining what distinguishes the behaviors that we call temperament from similar behaviors in other domains. Certainly, if we look at the major features of the working definition of temperament (early appearing, relatively stable, biologically rooted), these criteria easily could be applied to a variety of behaviors from distinctly different domains, such as intelligence or motivation. At the dimensional level, in Strelau's (1989) theory of temperament, activity level is linked to goal-directed behavior, with individual differences appearing in the difficulty and complexity of activities in which the individual engages. However, goal directed persistence and preference for challenging tasks have been claimed as fundamental aspects of mastery motivation (Barrett & Morgan, 1995). Similarly, the potential overlap between negative emotionality as a dimension of temperament versus negative emotionality as an index of internalizing behavior disorders has long been noted (Le-mery, Essex, & Smider, 2002). At the domain level individual differences in self-regulation are viewed as due to the operation of inhibitory control processes, which are linked to attentional mechanisms such as selective attention (Rothbart, 1991). However, inhibitory control has also been labeled as a cognitive phenomenon (Williams et al., 1999), while both attentional flexibility and attentional control (e.g., vigilance) also have been claimed as domains of information processing by cognitive theorists (Kinchla, 1992). Similarly, self-regulation itself is viewed by many cognitive theorists as the "hallmark" of cognition and cognitive processing (Borkowski & Dukewich, 1996).

It could be argued that the question of what field of study a given construct "truly" belongs to could be viewed as a form of semantic nit-picking. However, I see this problem as potentially much more fundamental. One of the major criticisms of social science theories is their lack of preciseness (Dar, 1987). I would argue that such imprecision is more likely to occur when theories that are designed to explain the nature of a particular domain (e.g., temperament) simply assimilate constructs from other domains as an essential part of the theory, without due regard for the origins of such constructs. For example, individual differences in activity level may reflect disorganized behavior as in the case of attention deficit hyperactivity disorder, in which case activity would serve as a risk influence upon subsequent development. Alternatively, individual differences in activity level may reflect temperament driven goal directed surgency, in which case activity could serve as a protective influence upon subsequent development (Wills et al., 2001). As seen in the case of classic psychoanalysis, theories that are definitionally imprecise ultimately become overelastic: being able to explain everything while truly explaining nothing. While current theories of temperament cannot be considered as overelastic, I believe there is a danger of these theories evolving in this direction, unless they attempt to deal with perennial questions such as what criteria can be used to define the realm into which a particular individual characteristic best fits. This problem is further compounded when researchers do not distinguish between different aspects of temperament dimensions (e.g., focused attention versus distractibility versus attention shifting) and lump them together as a composite index (Belsky et al., 2001). When this is done, specific predictive relations may be lost, as in the case where different dimensions of inhibition have been shown to be related to the ability to sustain ongoing behavior versus the ability to inhibit ongoing behavior (Kochanska, Coy, & Murray, 2001). To deal with such issues I would argue that it is essential to go beyond studying temperament in isolation and look at how and how much the various domains of temperament noted above overlap with similarly named constructs from other nontemperament domains.

There are two ways to deal with the question of potential overlap between similarly named constructs from different domains. As exemplified in our working definition, the first way is to acknowledge that we do have a construct definition problem. Among other things, this means not assuming that, because a trait fits our working definition of temperament, the trait is uniquely temperament in nature. Particularly for individual characteristics involving early personality traits, some individual traits do appear to fit better within the domain of temperament than in other domains. Traits fitting into this category would include emotionality, reactivity, difficultness, and activity. For other individual traits, such as attention or task orientation, where nontemperament domains appear to have equivalent claims, I would argue for viewing these as a separate "hybrid" class. Hybrid traits, while sharing some features with more classic temperament characteristics, also share features with characteristics from nontemperament domains. This situation is illustrated in Figure 2.1.

Viewing individual characteristics such as attention or task orientation as a hybrid class is not a semantic "sleight of hand" device. There are conceptual and analytic tools available that are designed to deal with situations where the boundaries that determine classification of a trait are unclear. Particularly for hybrid individual traits whose characteristics allow them to be assigned to any one of a number of domains, one potential approach is to apply what has been called "fuzzy logic" (Ohayon, 1999). Fuzzy logic involves both a theory and a method that can be used in situations where there are multiple, nondichotomous linguistic constructs that have overlapping boundaries (e.g., the boundaries of attention as temperament may well overlap with the boundaries of attention as cognition). Using fuzzy logic, it is possible to assess the degree of resemblance of a specific characteristic to the characteristics of a larger class, with the degree of resemblance ranging from zero (no overlap) to 1.0 (total overlap). For what appear to be nonhybrid traits, like emotionality or reactivity, we would expect the degree of resemblance between characteristics of these traits and characteristics of traits drawn from cognition or motivation to be essentially zero order. For what appear to be hybrid traits, like attention or persistence, we would expect the degree of resemblance to characteristics drawn from cognition or motivation to be more than zero order, but less than perfect in magnitude.

Figure 2.1 Example of a hybrid grouping of traits. Unshaded areas refer to trait characteristics that are unique to a specific domain. Shaded areas refer to aspects of trait characteristics that overlap multiple domains.

Use of fuzzy logic approaches would make it possible not only to determine to what degree a trait fits the criteria for membership in the domain of temperament versus the degree to which it is a hybrid trait, but also to determine what characteristics of the trait are more likely to fall under the membership criteria for each class. For example, for the hybrid dimension of attention, experts in cognition and temperament could be given lists of characteristics defining attention and their task would be to rate such characteristics as more likely to reflect attention as cognition or attention as temperament. Using fuzzy logic analytic procedures similar to those applied to understanding how individuals categorize objects or make social judgments (Massaro, 1987), such ratings could be used to determine which characteristics are most salient in defining the temperament components of attention versus those that are more applicable to attention as cognition. The former class of characteristics would be those we would want to emphasize when designing temperament assessment procedures that include attention as a salient dimension. Obviously, this is a very different line of research from what temperament researchers have done in the past. However, the use of alternative strategies, such as fuzzy logic, would seem to be a potentially important future line of inquiry if we are to truly understand what temperament means, and what defines those individual characteristics that uniquely or primarily fall within the domain of temperament.


Parent Agreement

As noted, one of our major assumptions underlying the construct of temperament is that temperament is a stable individual characteristic. Stability of temperament is assumed to occur both across situations and across time. Given the possibility of different contexts influencing children's expression of temperament at home and at school (Goldsmith, Rieser-Danner, & Briggs, 1991), it is perhaps not surprising to find only modest relations between parent and preschool teachers temperament ratings of individual children (Eisenberg et al., 2001; Lemery et al., 2002). However, contextual influences should be less likely to occur with mother-father ratings of their child's temperament, given that parent ratings presumably are based on observing the child in the same environmental context—the home. Thus, it is extremely troubling to find only moderate agreement by parents on the nature of their child's temperament (Mangelsdorf, Schoppe, & Buur, 2000; Slabach, Morrow, & Wachs, 1991), with mean correlations between mothers' and fathers' ratings of infants and toddlers (Goldsmith, 1996), preschool children (Lemery et al., 2002) or school age children (Rothbart et al., 2001) typically in the r = .3 to r =. 4 range. The modest levels of interparent agreement shown for ratings of children's temperament are problematical, if for no other reason than their implications for the stability of temperament across time. How can we propose individual stability of a child's inhibition or activity level across time when there is disagreement at the initial measurement point about how temperamentally inhibited or active a child actually is?

What might be the reasons for the moderate level of parental agreement about the characteristics of their child's temperament? The hypothesis that fathers' ratings are less accurate because fathers spend less time with their children has not received strong empirical support (Slabach et al., 1991). There is some evidence indicating that level of agreement between parents varies as a function of the temperament dimension being considered. Higher interparent agreement has been found to occur for scales assessing negative emotionality (e.g., fear, fussy, difficultness), activity or shyness than for scales assessing positive emotionality, attention, perceptual sensitivity or persistence (Goldsmith, 1996; Goldsmith & Campos, 1990; Mebert, 1989; Rothbart et al., 2001). These data suggest that parents may be more attuned to those aspects of their child's temperament, such as high activity, negative emotionality or shyness, that call for more active parental involvement (Bell & Chapman, 1986). However, even for these domains, the level of parental agreement rarely goes beyond the r = .5 range, suggesting that other factors are operating besides just some temperament dimensions be ing more salient for parents (Rothbart et al., 2001). With regard to the possibility that the structure of temperament may differ for mothers and fathers (e.g., mothers and fathers may be using different behaviors to represent the same temperament dimension for their child), the evidence is mixed. When subscales from temperament questionnaires are the basis of analysis some studies show different factor structures underlying mothers' and fathers' ratings of their child's temperament (Goldsmith & Campos, 1990). In contrast, other studies suggest little difference in the types of behaviors used by mothers and fathers to rate specific temperament characteristics of their child (Hubert, 1989). The extent to which such differences reflect methodological differences among studies in regard to what parents are rating remains unclear.

It is possible that the moderate level of interparent agreement is a narrow methodological issue that is unique to questionnaire measures of temperament, perhaps reflecting the possibility that parental temperament ratings primarily reflect subjective aspects of parental rather then child characteristics (Mangelsdorf et al., 2000). Differences in mothers' and fathers' ratings of their child's temperament have been associated with levels of parental anxiety or depression (Mebert, 1991). However, such a conclusion is problematical given evidence that parental report measures of their child's temperament are not purely subjective, but also assess real existing differences in those domains of individual child characteristics which we call temperament (Bates, 1994). For example, if parental report measures of their infant's temperament were purely subjective, we would not expect to see the significant levels of relation between parent report measures and objective laboratory assessments of temperament that repeatedly have been documented in the literature (Rothbart & Bates, 1998; Slabach et al., 1991).

If the moderate levels of interparent agreement documented above cannot be understood solely by reference to issues involving the measurement of temperament per se, what other possibilities are there? Continuing the overall theme of this chapter I would argue that we need to go beyond temperament in isolation if we are to more fully understand the sources of moderate levels of reliability of parent temperament ratings. One possibility is that nontemperament individual child characteristics may play a role in explaining the modest level of interparent agreement. Some evidence has suggested that mothers' ratings are influenced more by the age of their child while fathers' ratings tend to be influenced more by their child's gender and parity (Slabach et al., 1991). Another potential explanation is based on the possibility that while parents share the same overall home context with their children there may be differences in the types of behaviors that mothers and fathers use when interacting with their children, and that these differences may be reflected in different temperament related behavioral patterns displayed by the child to their mother and father (Mangelsdorf et al., 2000; Rothbart et al., 2001). Unfortunately, at present all too little evidence is available relating differences in parental ratings of child temperament to differences in child characteristics or to differences in parent-specific interactions with their children.

Reviews focusing on biological contributions to temperament have offered another potential explanation for the modest levels of interparent agreement. Both Nelson (1994) and Rothbart, Derryberry, and Posner (1994) have noted two fundamental points about central nervous system contributions to the development and expression of temperament. First, there are multiple central nervous system areas involved in the development and expression of temperament. For example, behavioral inhibition appears to develop as a function of coordinated interaction between the hippocampus, prefrontal cortex, and portions of the motor system. Second, the different central nervous system areas which underlie the development of specific dimensions of temperament reach functional maturity at different times. For example, the hippocampal region matures earlier than does the prefrontal cortex (Nelson, 1994). What are the implications of these biological foundations for understanding levels of interparent agreement? As pointed out by Nelson (1994), while young infants have the capacity for short-term reactivity to stimulation, they may not have the capacity during the first year of life to store emotional memories associated with such reactivity. What this means is that the developing central nervous system structure of young infants may not permit consistency across different situations, until the relevant central nervous system structures that allow emotional memory storage have reached functional maturity (sometime between 12 and 24 months). As a result, mothers and fathers may be seeing fairly inconsistent reactions from their young child to particular stimuli, making it difficult for them to come up with a coherent and agreed on characterization of their infant, particularly if mothers and fathers treat their infant differently (Wachs & King, 1994). This does not mean that the young infant is deliberately tailoring different reactivity patterns to their mothers and their fathers, but rather that the nature of the early central nervous system may not permit infants to display consistent emotional reactivity patterns. Indirect support for this hypothesis is seen in the increase in parental agreement across the first two years of life (Mangelsdorf et al., 2000) with little change in level of agreement between the preschool and early school years (Rothbart et al., 2001). Such differences could reflect the faster rate of brain development earlier in life, as compared to the slower developmental rate after preschool. Again, this neurobiological approach to the question of moderate interparent agreement is one that could not have been developed by focusing on the measurement of temperament in isolation.

Getting to Know Anxiety

Getting to Know Anxiety

Stop Letting Anxiety Rule Your Life And Take Back The Control You Desire Right Now! You don't have to keep letting your anxiety disorder run your life. You can take back your inner power and change your life for the better starting today! In order to have control of a thing, you first must understand it. And that is what this handy little guide will help you do. Understand this illness for what it is. And, what it isn't.

Get My Free Ebook

Post a comment