Prepared as a chapter for the Annual Review of Psychology, 1995.
For more information on personality theory and research, go to the Personality Project

SOME PEOPLE ARE THE SAME: THE STUDY OF INDIVIDUAL DIFFERENCES AND SIMILARITIES

That people who share a similar upbringing are more similar than those who do not is obvious. Similarities based on linguistic and cultural background have never been denied. The utility of using individual differences in one situation to predict individual differences in another situation, however, has been hotly contested. Debates about the relative importance of situational versus individual causes of consistencies and differences dominated a disproportionate amount of the literature of the 1970's and 1980's but became less virulent as both sides developed more tolerance, became exhausted, or developed a richer understanding of the underlying issues. By changing their emphases, both sides have made theoretical advances by better understanding their limits.

Trait based differences and similarities

In a tradition strongly associated with prediction and selection (Kanfer et al in press), the study of individual differences in personality represents the greatest amount of personality research. Indeed, so much work has been done that to some the field of personality is the study of individual differences (A. Buss 1989). Just as personality theories can be organized in terms of their level of explanation and level of generality, so can studies of individual differences be further organized along two dimensions: cognitive vs. affective-temperamental and descriptive vs. causal-explanation.

Cognitive abilities versus affective-temperamental traits

The first dimension distinguishes analyses of intellectual abilities from those of non-cognitive variables associated with affective reactions and behavior. The cognitive/non-cognitive distinction runs throughout the field and some personality theorists specifically rule out cognitive ability as an area of study. Others include both cognitive abilities and temperamental traits as part of personality structure, whereas still others discuss personality and intelligence as separate domains but routinely study both. The distinction between cognitive and affective components of personality is both clarified and muddied by the labels given to measures in these two domains: tests of intellectual ability and tests of personality.

Ackerman and Kanfer and their colleagues have attempted tointegrate individual differences in cognitive and non-cognitive function in applied settings. Not only do cognitive and non-cognitive measures differ in content, but they also differ in typicality. Intelligence tests are meant to be maximal performance measures whereas most non-cognitive scales are measures of typical behavior. Furthermore, cognitive tasks are direct behavioral measures whereas non-cognitive measures are typically based on self reports of average behaviors or of intentions (Brody 1994). The predictive relationship with performance outcome of cognitive and non-cognitive measures changes as people become more experienced with the task at hand. Non-cognitive measures and typical intellectual engagement become more important predictors over trials and maximal cognitive performance becomes less important (Ackerman, in press). Further clarifying the relationship between maximal and typical performance, Goff & Ackerman (1992) report that typical intellectual engagement, although independent of fluid intelligence, is correlated positively with crystallized intelligence. Typical intellectual engagement is highly related to the "Big 5" dimension of openness (Rocklin 1994) but differs somewhat at lower order components of both openness and typical intellectual engagement (Ackerman & Goff 1994).

An innovative use of cognitive and non-cognitive variables in the study of creativity is proposed by H. Eysenck (1993). Eysenck reviews the J shaped distribution of creative output and suggests that it results from the interactive product of cognitive ability, societal constraints and opportunities, and non-cognitive variables including confidence, non-conformity, and originality. He places particular emphasis upon the role that the psychoticism dimension plays in creative productions.

Experimental analyses have shown systematic although complex relationships between non-cognitive personality variables and cognitive performance. These relationships are moderated by a variety of situational manipulations that affect motivational states (Anderson, in press; Anderson & Revelle, in press; Matthews et al 1990; Matthews et al 1989; Revelle 1989). These are theoretically driven tests of the arousal model of extraversion (H. Eysenck 1967, 1990) and its modification and extension to impulsivity (Humphreys & Revelle1984). Helpful reviews of the effects on performance of extraversion (Matthews1992a) and anxiety (Mueller 1992) summarize many theoretical approaches to the combination of cognitive and non-cognitive individual differences.

In applied settings, cognitive measures have been used since at least the Army Alpha test in World War I. Non-cognitive variables have a long and checkered past but "the emergence of an acceptable taxonomy of personality during the 1980s has provided applied psychologists with a sorely needed organizational framework for investigation of personality-work linkages." (Kanfer et al in press p 30). Conscientiousness, experience, and ability combine to predict job performance (Schmidt & Hunter 1992).

Descriptive taxonomies versus causal theory

After many years of bitterly fought debate about the appropriate number and identification of the fundamental dimensions of personality, the past several years have seen a remarkable consensus among most but not all descriptive taxonomists around five robust factors (the "Big 5" or B5): Extraversion, Emotional (in)Stability or Neuroticism, Agreeableness, Conscientiousness, and Openness or Culture. (See Digman 1990, Goldberg 1992, 1993a,b, John 1990, Ozer & Reise 1994, Widiger & Costa 1994, Wiggins & Pincus 1992). Critics have suggested that this consensus is premature and overstated (Pervin 1994). Although much of the work on the B5 addresses the number and identification of personality dimensions, there are some particular instantiations such as the Five Factor Model (FFM) that are more concerned with underlying mechanisms (John & Robins 1993, McCrae & Costa 1990, in press).

Descriptive taxonomies of individual differences have been a tradition in personality theories since Plato and Galen. Most taxonomic systems of cognitive and non-cognitive attributes are hierarchical: clustering similar behaviors into narrow traits, then clustering these into higher order traits, and eventually into a limited number of dimensional types (H. Eysenck 1991a). At any level of this hierarchy, behaviors and traits can be found that represent blends of separate dimensions, resisting any appearance of factorial simple structure and requiring a horizontal as well as a vertical structure (Goldberg 1993a,b). The problem for taxonomists thus becomes determining the optimal number of factors to describe these structures. Optimality means different things to different investigators, but includes being parsimonious, replicable, and useful. It is not surprising that there is not perfect agreement among all taxonomists given the many assumptions implicit to factor or principal components analysis.

There is strong agreement that the dimensions of extraversion/introversion and neuroticism/emotional stability are fundamental parts of any personality taxonomy. But proponents of what can be called "The Even Bigger 3" (EB3) suggest that openness is more of a cognitive than non-cognitive construct, and that agreeableness and conscientiousness are both parts of a higher order factor of Psychoticism (H. Eysenck 1990, 1991b), or Psychoticism-Impulsivity-Sensation Seeking (Zuckerman 1991, 1994).

The dimensions of the B5 and the EB3 can be used to classify and provide order to the multiplicity of psychiatric diagnoses found useful by therapists and clinical researchers. The numerous personality disorders listed in the Diagnostic and Statistical Manual of Mental Disorders (DSM IIIR) (American Psychiatric Association 1987) may be organized parsimoniously in terms of the FFM (Widiger & Costa 1994). Neuroticism is a risk factor for depression and anxiety, and introversion in combination with neuroticism increases risk of depression (Clark et al 1994; Costa & McCrae 1993). The hypothesized biological basis of the EB3 has been used to organize both the Axis I and Axis II dimensions of the DSM IIIR (Siever & Davis 1991). A re-conceptualization of the EB3 has been used to categorize the personality disorders (Cloninger 1987). It is uncertain whether personality traits are the causal sources of psychopathology, co-occurring signs, or the resulting psychological scars left by experiencing these disorders. Taxonomic work on interpersonal problems suggests a general factor of distress and a two dimensional—circumplex structure that has been described in terms of love and trust or the B5 dimensions of emotional stability, extraversion, and agreeableness (Gurtman 1992, in press).

Taxonomic studies of individual differences in mood have extended the earlier work of Tellegen (1985), Russell (1979); Thayer (1989), and Watson & Tellegen (1985) on identifying two independent dimensions of mood and emotion that are associated with positive and negative affect or energetic and tense arousal. These two dimensions of mood are, in turn, related to the EB3 and the B5. Extraversion is associated with measures of positive affect, neuroticism with measures of negative affect (Meyer & Shack 1989; Saucier 1992; Watson et al 1994).

Causal models of individual differences

The descriptive taxonomies associated with the proponents of the B5 are in contrast to those theories concerned with developing causal models of individual differences. Much of the recent consensus around the B5 has been on the number of dimensions useful in the description of individual differences rather than in any causal basis for these purported structures. Descriptive taxonomists suggest that before it is possible to develop causal explanations it is necessary to agree on the fundamental dimensions to be explained. Causal theorists, on the other hand, have focused on biological explanations of the EB3 and have emphasized the relationships of biological mechanisms of emotional reactivity with dimensions of stable individual differences. These theorists have suggested that problems of taxonomy can best be solved in terms of underlying mechanisms.

Until recently, this work has followed two related paths: demonstrations of the genetic basis of particular traits and explorations of particular biological mechanisms thought to be associated with individual differences in the major affective and cognitive traits (see D. Buss 1990). Among the mechanisms proposed are differences in relative activation of specific brain structures as well as differences in the relative amounts of specific neurotransmitters. More recently, some causal explanations for individual differences of some of the B5 have been proposed in terms of evolutionary theory (MacDonald 1992).

Genetics of individual differences in behavior

Recent evidence suggests that practically any trait of interest has a substantial genetic component. Excellent monograph length reviews summarize studies of the heritabilities of the B5 (Loehlin 1992), review how behavior genetic studies help clarify the "nature of nurture" (Plomin 1994) or integrate many different approaches (Plomin & McClearn 1993). Behavioral patterns as complex as sexual orientation for males (Bailey & Pillard 1991) and females (Bailey et al 1993), political attitudes (Tessor 1993) or various personality disorders (Nigg & Goldsmith 1994) show strong evidence for heritability. Direct comparisons of results from cognitive and non-cognitive studies suggest higher heritabilities of cognitive measures than of B5 or EB3 non-cognitive dimensions (Brody 1993, 1994), although this may be the result of differences in scale reliability.

Kimble (1993) points out that the sudden resurgence of debates about nature versus nurture is surprising for those who remember learning that asking which is more important, is like asking which contributes more to the area of a rectangle, the width or the length. That almost all of the major personality dimensions seem to have a substantial (50% +/- 20%) heritability is no longer a point of contention. What is more interesting is the genetic nature of the covariances between traits, the way in which one's environment contributes the remaining 30-70% of the variance (Brody 1993; Plomin 1994), and thus, the way that genotypes lead to phenotypes.

Heritability is a population value reflecting the amount of between individual variability associated with additive (narrow heritabilty) or total (broad heritability) genetic variation. Heritability does not imply immutability; a lack of heritability does not imply a lack of a biological mechanism; nor does a high heritability imply a simple biological mechanism. Thus, that 90% of the variance in height is under additive genetic control does not preclude a several centimeter increase in height due to improved nutrition. That there is no additive genetic component to the sex of one's offspring does not imply that sex chromosomes don't matter. And finally, that 52% of the variability in the likelihood of divorce is under genetic control (McGue & Lykken 1992) does not imply that there is a divorce gene, nor does it imply that there is a divorce nucleus somewhere in the limbic system, nor is it inconsistent with large temporal variation in divorce rates. Complex behaviors reflect the sum and interaction of many separate predispositions. Although the similarity of monozygotic twins reared apart allows for estimates of broad heritability (Bouchard et al 1990; Bouchard & McGue 1990), these estimates may be inflated estimates of narrow (additive) genetic influences due to the effects of scaling, dominance, and gene-gene interactions (epistasis, also called emergenesis by Lykken et al 1992).

Creative research designs take advantage of the power of structural modeling procedures to estimate genetic and environmental parameters from adopted and biological siblings living together, biological siblings living apart, parent-child correlations for adopted and biological children, and many of the other living arrangements modern society provides (Eaves et al 1989; Loehlin 1992; Plomin et al 1990). Each unique family constellation can be fitted with alternative genetic models and the resulting path diagrams allow one to choose the most parsimonious.

One consequence of systematic modeling is that much more is known about environmental influences on personality development than was known before behavior genetic modeling was done. Just as classic behavior genetics analysis allows for a decomposition of genetic variance into additive and non-additive within and between family genetic effects, so can estimates be derived for shared and unshared family environment effects. A striking conclusion is that in general, the shared family environment contributes little if anything to the similarity of children growing up in the same home. That is, within a similar culture, biologically related children growing up together tend to be as similar on most personality traits as they would growing up apart, and unrelated children growing up together tend to be no more similar than unrelated children in general. In fact, it is likely that some similarities of child rearing practices that are experienced by children are largely the result of the genetic similarity of the children. Environment means more than one's family, for it includes cultural as well as prenatal environments.

Behavior genetic techniques can be applied to the covariances between traits as well as the variances of traits. These techniques, although long available, have become more useful as larger samples and more powerful algorithms have become available. A useful example of such modeling is the examination of the genetic covariance of neuroticism, anxiety, and depression to analyze the direction of causation between them (Carey & DiLalla 1994).

The consistency of the behavior genetic evidence can be interpreted in two ways: Rather than showing whether or not environments are important determinants of personality, genetic modeling has shown how dynamic is the process of personality development. Gene-environment covariation suggests that people are selecting and shaping the environments in which they live, rather than being passively acted upon by the environment. Children shape the action of their parents just as parents try to modify the behavior of their children (Rowe & Waldman 1993; Scarr 1992).

Biological substrates

Genes do not act directly on behavior. Genes code for proteins that in turn affect structures and regulate processes. Most biologically based theorists have asked what particular structure, neural pathway, transmitter, or hormone is associated with a particular individual difference in affect, cognition, or behavior. Much of this theorizing has been at the level of the conceptual nervous system (cns) rather than actually describing the Central Nervous System (CNS). That is, broad brush behavioral systems have been described and linked, sometimes closely, sometimes loosely to known physiological structures and transmitters. To the biologically oriented radical trait theorists, taxonomies should be developed in terms of cns or CNS biological systems rather than phenotypic behaviors. Individual differences in the functioning of these systems are believed to cause differential sensitivities to environmental cues, leading to differential affective and cognitive states. Traits refer to the probabilities of being in a particular state, or to the latency to achieve a state following a specific environmental elicitor. Although it is not necessary to know the specifics of a neural system to test the implications of a conceptual system, by limiting theorizing to known neural architectures, personality theories become more constrained.

That the proposed biological mechanisms for these conceptual systems differ from investigator to investigator should not be taken as a sign of theoretical weakness but rather a sign of the complexity of the purported systems. No single structure, transmitter, or gene controls the entire system, but rather each plays a supporting and limiting role. Consider by analogy the case of oxygen flow to the brain. Experimental demonstrations of the importance of the heart to oxygen levels in the cortex or low correlations across subjects between measures of heart and lung functioning do not imply that the lungs, veins, arteries, and vagal nerve are not also involved in oxygen transport and regulation, for they are all in fact part of the same circulatory system.

Further complicating any simple review is the multiplicity of analytic techniques. Data are reported in terms of structures, transmitters, and electrophysiology. Generalizations are drawn from rodents, primates, and humans. Dominant EEG frequencies for children are labeled in terms of higher frequencies found in adults. Correlations are made with structures identified by MRI and PET, or with functioning observed by evoked potentials or hormone levels.

Most experimental and theoretical statements concerning the biological substrates of personality are directly or indirectly related to the theories of Hans Eysenck, whose theory of the biological basis of introversion-extraversion, neuroticism-stability, and socialization-psychoticism (H. Eysenck 1990) has evolved from taxonomic work (H. Eysenck 1947) to a proposed biological model (H. Eysenck 1967) that has been the basis of a variety of suggested modifications (Cloninger 1987; Gray 1972, 1981, 1991, 1994). In broad strokes, Eysenck's theory and subsequent modifications (1990, 1991a) are theories of approach and reward, inhibition and punishment, and aggression and flight. All three constructs have been, of course, fundamental concerns for many years and have been the basis for descriptive as well as non-biological theories of motivation and learning (Atkinson 1960; Dollard & Miller 1950). Approach and withdrawal are behavioral characteristics of amoebae, insects, and human infants (Schneirla 1959). Unifying recent biological work is an emphasis on these three interrelated biological and behavioral systems as sources of individual differences in affective reactions and interpersonal behavior. Although differing in the particular mechanisms proposed at the level of the CNS, these models show striking agreement at the behavioral and conceptual (cns) level.

Central constructs of Eysenck's biological theory of introversion-extraversion (I-E) and stability-neuroticism were cortical arousal and limbic activation (H. Eysenck 1967). Arousal was originally postulated as reflecting activation of the Ascending Reticular Activating System (ARAS) and the associated cortical-reticular loop. Activation reflected limbic activity of the hippocampus, amygdala, singulum, septum, and hypothalamus. Introverts were thought to have higher levels of resting arousal than did extraverts. With the assumption that some intermediate level of arousal was preferred, the stimulus seeking behavior of extraverts was explained as a compensation for a lower resting level. With the recognition that ARAS arousal was too broad a concept, Eysenck subsequently modified his theory to include a limbic arousal system, the monoamine oxidase system, and the pituitary-adrenocortical system (1990). He suggested that the apparent diversity of multiple arousal mechanisms "may not prevent the systems from operating in a relatively unified fashion." (p 249). He associates subjective arousal with Thayer's (1989) measures of energetic arousal (feelings of energy, vigor, and pep). Extraversion and feelings of energetic arousal are both associated with approach behavior and with positive affect following reward or cues for reward.

Stelmack (1990) summarized 20 years of psychophysiological research on Eysenck's hypothesis that introverts have higher arousal levels than extraverts and concluded that there "is a good deal of evidence that introverts are characterized by greater physiological reactivity to sensory stimulation than extraverts ... [but] there is little compelling evidence that introverts and extraverts differ in tonic or basal levels" (p 307). Indirect tests of the arousal hypothesis have examined the relationship between extraversion and cognitive performance under various experimentally induced arousers. These studies do not support the hypothesis of a stable I/E difference in tonic arousal levels, but they do show that I/E, or the impulsivity component of I/E, moderates the relationship between induced arousal and performance (Anderson, in press; Anderson & Revelle, in press; Matthews 1992a; Matthews et al 1989, 1990; Revelle 1993).

The Behavioral Approach System (BAS, Gray 1994), also known as the Behavioral Activation System (Fowles,1988), or Behavioral Facilitation System (Depue & Iacono 1989), activates approach behaviors in response to cues for reward or non-punishment. It may be associated neurophysiologically with the motor programming system. "The key components are the basal ganglia (the dorsal and ventral striatum, and dorsal and ventral pallidum); the dopaminergic fibers that ascend from the mesencephalon (substantia nigra and nucleus A 10 in the ventral tegmental area) to innervate the basal ganglia; thalamic nuclei closely linked to the basal ganglia; and similarly, neocortical areas (motor, sensorimotor, and prefrontal cortex) closely linked to the basal ganglia." (Gray 1994, p 41). Dopamine is said to play an essential moderating role in the functioning of the BAS (Depue & Iacono 1989; Depue, et al., in press), but the full relationship of dopaminergic activation and reward is less than clear (Wise and Rompre 1989).

The cluster of approach traits of extraversion (H. Eysenck 1990), impulsivity (Barratt, in press; Gray 1994; Zinbarg & Revelle 1989), novelty seeking (Cloninger 1987), and positive affectivity (Depue & Iacono 1989; Depue et al in press; Tellegen 1985) as well as the states resulting from approach or reward, energetic arousal (Thayer 1989), and positive affect (Watson et al 1994) have all been discussed in terms of the BAS.

If the BAS is the engine of behavior, the Behavioral Inhibition System (BIS) is the braking system. Signals of punishment, non-reward, novel stimuli, and innate fear stimuli lead to behavioral inhibition, an increment in tense arousal, and increased attention. The BIS may be considered both as a cognitive and physiological system (Fowles,1988; Gray 1982). Cognitively, the role of the BIS is to compare the current state of the world with expectations, and to inhibit and modify behavior that leads to deviations from expectation. Physiologically, the comparator function of the BIS is associated with the septohippocampal system. Input to this system comes from the prefrontal cortex, and output flows through the noradrenergic fibers of the locus coeruleus, and serotonergic fibers from the median raphe (Gray 1994). More detailed reviews of the neurophysiology of the BIS emphasize the role of seretonin (Clonginger 1987, Depue & Iacono 1989, Spoont 1992) and the amygdala (Kagan, Arcus & Snidman, 1993).

Just as approach traits are associated with the BAS, so are avoidant and inhibitory traits associated with the BIS. Anxiety and neuroticism are believed to reflect chronically high levels of BIS function (Gray 1994). Negative affect and state anxiety are both state markers of BIS activation. Depression has been proposed to reflect high BIS and low BAS activity (Clark & Watson 1991).

Aggression and hostility have been associated with the third dimension of the EB3, psychoticism. In terms of the B5, psychoticism is a combination of (dis)Agreeableness and (un)Conscientiousness. Neurologically, the Fight Flight System has been associated with the amygdala, the medial hypothalamus, and the central gray of the midbrain (Gray 1994). Neurochemically, serotonin, gamma-aminobutyric acid, the endorphins, and testosterone have been implicated in aggression and hostility (Dabbs & Morris 1990).

In partial agreement with the dichotomization of affect into positive and negative systems are studies of the lateralization of emotionality that suggest an association between left frontal activation and approach-related positive affect and right frontal activation and inhibitory or withdrawal related behavior and negative affect (Davidson 1992, 1993a, in press). Unfortunately, identification of particular biological systems with particular personality traits or psychopathological disorders tends to ignore the complexity of neural architecture. The brain has evolved to solve many different problems and primitive systems are controlled by later, more complex systems (Derryberry & Tucker 1992; MacLean 1990).

Amelang & Ullwer (1991) and Fahrenburg (1991) discuss data that are quite critical of the uniform acceptance of simple relationships between self report dimensions and biological systems. Their criticism is two fold: the complex specification of parameters necessary to find the purported results make theories overly complicated, and based upon their empirical investigations, there is little evidence for the proposed mechanisms.

The need to optimally specify parameters to detect presumed relationships has long plagued the personality research (H. Eysenck & Levey 1972). It is useful to consider this issue in some detail, for appropriate parameter values are a consistent difficulty in personality research. At the most naive level, individual differences in a trait would be expected to produce consistent individual differences in an associated construct independent of other parameters. This assumption is typical of classical test theory, which assumes that equal true score differences are equally discernible (i.e., will lead to equal observed score differences) at all levels of true score. More recent approaches (e.g., item response theory), however, make it clear that this is not the case. Observed scores are a monotonic, but non-linear function of individual differences on some underlying attribute. They also reflect differential response probabilities (difficulty) associated with the situation (item) being measured. That is, even large differences on a latent trait are not easily detectable if the situational parameters are inappropriately specified. Such scaling artifacts include ceiling or floor effects. These scaling problems can lead to inappropriate inferences about group differences as well as about interactions of multiple variables (Revelle & Anderson 1992).

Another problem arises when consistent individual differences reverse direction depending on the value of a specific parameter. This is not a problem of mere scaling but can be a much more serious theoretical challenge. Some reversals are predictable consequences of non-monotonic relationships (Anderson 1990, 1994; Humphreys & Revelle 1984; Yerkes & Dodson 1908). But other reversals can not be explained in terms of theory related parameters. For example, the supposed greater rate of decay in performance over time for high impulsives than for low impulsives reverses when studies are conducted in the morning versus the evening (Anderson & Revelle, 1994).

Temperament and development

In parallel with the development of the biological models of personality developed from rodents and adult humans is the work on children. As all parents know, children are different from each other. Some are shy, some are bold, some are slow to warm up, some are unafraid of new challenges. Child developmental research concerned with seemingly biological traits has emphasized the temperamental aspects of personality development. This work on temperament has, until recently, been somewhat independent of the adult research literature in personality, although "a complete understanding of personality and psychopathology must be a developmental one" (Rothbart & Ahadi 1994, p 55). Theories of adult personality and childhood temperament, besides being isolated from each other, have tended to be parochial, with a lack of communication between American and Eastern European researchers. Attempts have been made to reverse both of these trends, with conferences and edited volumes reporting work on both adults and children with contributions by both eastern and western researchers (Bates & Wachs 1994; Strelau & Angleitner 1991).

Temperament may be seen "as constitutionally based individual differences in reactivity and self regulation, influenced over time by heredity, maturation, and experience." (Rothbart & Ahadi 1994, p 55). Aspects of reactivity include activation of physiological and behavioral systems. Regulatory processes that modulate reactivity include selective attention and processing of cues to reward and punishment, as well as approach and inhibition to novel stimuli (Rothbart et al 1994). Discussions of reactivity and regulatory processes bare a striking resemblance to those of approach and avoidance traits in adults (Strelau 1987, 1994).

Temperamental differences in reactivity to novelty and to strangers (Kagan et al 1992) show striking correlates with hemispheric differences in activation (Davidson 1993b). Shy or inhibited children identified at 31 months showed more right hemispheric activation than did uninhibited children when they were tested at 38 months. This effect seems to be due to a deficit in the left-frontal approach system for the inhibited children rather than to a hyperactivation of the right-frontal withdrawal system (Calkins & Fox 1994; Davidson 1993a). Gray's model of approach and avoidance is discussed in the context of infants and the neural structure of temperament (Nelson 1994; Strelau 1994). In a discussion of the relationship between temperament and attachment, Goldsmith & Harman (1994) point out that physiological measures do not explain temperament and suffer the same difficulties in interpretation as do behavioral measures.

Longitudinal studies of temperamental differences suggest long term consistencies in behavior (Caspi & Bem 1990). In a continuing study of the antecedents and correlates of delinquency, B5 measures were found by Robins et al (1994) to be related to dynamic conceptions of ego-control, and ego-resiliency. In a 15 year long study of impulsivity and disinhibitory behavior, children diagnosed as hyperactive or attention deficit disordered continued to show impulsive behavior in early adulthood and to be at greater risk for alcoholism and committing violence (af Klinteberg et al 1994).

Affective and cognitive processes--how traits relate to states

Traits are not behavior. They are summary statements describing likelihood of and rates of change in behavior in response to particular situational cues. In addition to their relationship to the probability and latency of response, stable predispositions may be conceptualized in terms of differential sensitivities to situations and differential response biases. Intervening among traits, situations, and responses are momentary affective and cognitive states.

Taxonomic analyses of mood and emotion disagree about categorical versus dimensional representations. Do the many separate emotional terms in the natural language describe many different emotions, or are there a limited number of affective states that differ in intensity and duration? Two affective dimensions that relate to stable personality traits are positive and negative affect (Meyer & Shack 1989; Watson et al 1994) or the related constructs of energetic and tense arousal (Thayer 1989). Extraversion tends to be related to positive affect, neuroticism with negative affect. These relationships, however, are not strong and interact predictably with the situation. In positive, rewarding situations, extraversion is associated with positive affect, but this relationship vanishes in threatening situations. Similarly, neuroticism is related to negative affect under threat, but not under reward conditions (Larsen & Ketallar 1989). Although these relationships are consistent with theories of traits and states, they are small enough to require assessing traits and states separately in order to study relationships with performance (Matthews 1992b).

Further complicating the trait-state relationship is its dynamic nature. When free to choose situations, individuals sensitive to negative affect (neurotics) will try to avoid threatening situations. It is the emotionally stable individual who is more likely to participate voluntarily in activities that are likely to induce negative affect. Thus, it is necessary to distinguish between externally imposed and freely selected situations as well as within and between subject differences in the use of affective scales.

Traits as well as emotional states affect the detection, encoding, storage, retrieval and integration of information (Christianson 1993). Trait and state effects may be seen at each of these conceptual stages (Revelle 1993). Impulsivity interacts with time of day to affect energetic arousal, which in turn is related to the detection and storage of information (Anderson & Revelle, in press; Revelle & Loftus 1993). Anxiety shifts attention to threat related cues whereas depression biases memory towards depression-related material (Mathews 1993; Mueller 1992). Relations between anxiety and memory vary as a function of trait and state anxiety as well as implicit and explicit memory conditions (M. Eysenck & Mogg 1993).

Life satisfaction, identity, and death

Personality characteristics of young adults predict life long risks for neuroticism, and emotional health, and even death (McCrae 1994). The effect of neuroticism and extraversion on psychological distress over a ten year period has been estimated to be four times greater than the effects of psychological interventions to reduce distress (Brody 1994).

Conley (1985) examined the multitrait-multimethod-multitime structure of self reports and peer ratings of neuroticism, social extraversion, and impulse control from the Kelly longitudinal study measures taken in 1935-1938, 1954-1955, and 1980-1981. Neuroticism at times 1 and 2 reliably predicted neuroticism and emotional health at time 3. Similarly, social extraversion measures at times 1 and 2 predicted social extraversion and social activity at time 3.

Measures of conscientiousness taken in 1922 as part of the Terman (1925) study predicted mortality risk through 1986 through age 76 with a relative hazard of death of roughly .75. Stated differently, for someone at the 25th percentile of conscientiousness at age 12, there is a 35% greater risk of dying before age 70 than for someone at the 75th percentile (Friedman et al 1993).

Situational based differences and similarities

Social learning theory explains consistent individual differences in behavior in terms of stabilities in the supporting environment rather than in terms of an individual's characteristics. Consistency across situations reflects similarity of situations rather than stable individual traits. Behavior can be modified by changing the environmental cues. Total reliance on prior learning experiences rather than on individual readiness is as much a straw man for social leaning theory as total cross situational consistency is for trait theory. Adherents of social learning theory now emphasize the need to understand how individual cognitive representations of the environment lead to behavior (Cantor 1990; Cantor & Zirkel 1990).

Some of the clearest evidence for the effect of the formative and sustaining environment on determining individual differences comes from behavioral genetic analyses. That identical twins are not perfectly concordant for extraversion, neuroticism, schizophrenia, or homosexuality demonstrates environmental effects. More importantly, that identical twins growing up together seem to be no more similar than those growing up apart (Eaves et al 1989; Tellegen et al 1988) implies that the formative environment is not the set of experiences shared within a family, but is either unique to each individual or common to their culture.

Part of the unique family environment is birth order. Although genetically related, siblings differ in age, experience, and in reproductive value to their parents. Differences between siblings growing up together can be magnified by contrast effects. Sulloway applies an evolutionary perspective on sibling rivalry in a meta-analysis of birth order effects on the traits of the B5 and reports that first borns are more extraverted and conscientious but less emotionally stable, agreeable, or open than are later borns. Later borns are more likely to adopt radical innovations in science than are first borns (1995).

Trait by Situation Interactions

Although "interactionism" was claimed to be the new and improved way to study personality (Magnusson & Endler 1976) most personality research has gone beyond the simple assertion that consistencies exist in the interactions of traits and situations. Theoretically driven trait theorists have long recognized that stable individual differences produce predictably different patterns of results in different situations. Failure to change one's actions across situations is a sign of pathology, not adaptive behavior. The utility of demonstrations of trait by situation interactions lies in the exclusion of many competing hypotheses, as well as in the setting of boundary conditions for individual and situational effects, for theories are best tested at their limits.

Consider delinquency as an example of the setting of boundary conditions. Delinquency may be conceived as the outcome of the interaction of lack of social constraints with a biological propensity. When social constraints are diminished, the relationship between testosterone and delinquency and antisocial behavior increases (Dabbs et al 1990; Dabbs & Morris 1990). Among lower SES military veterans there is a positive relationship between testosterone and antisocial behavior, but this relationship vanishes among higher SES subjects. These results might be due to a lack of social control, or alternatively, to the existence of more legal ways to seek stimulation among higher SES groups than among lower SES groups.

Interactions also allow for tests of theories. Consider the relationship between impulsivity and cognitive performance which changes as a function of caffeine (Anderson, in press) or time of day (Anderson & Revelle, in press) and differs as a function of the particular task used (Revelle et al 1987). These interactive results allow for precise tests of the competing theories relating to the arousal interpretations of impulsivity (Revelle & Anderson 1992).

In a thoughtful review of the many meanings of person x situation interactions, Higgins (1990) emphasizes the interplay between situational standards and individual beliefs. Differences in cognitive representations and activation prime reactions to specific situational cues. This social psychological emphasis upon contextual priming of memories relates to the personality concern with individual differences in cognitive structures. Continue

Prepared as a chapter for the Annual Review of Psychology, 1995.
For more information on personality theory and research, go to the Personality Project