[1] Much of the work that I report has been done in collaboration with colleagues and students of mine. Most importantly, I want to recognize the contributions of my former colleague Michael Humphreys (now at the University of Queensland) and my former student and current colleague Kristen Anderson. In addition, the research of recent Ph.D. students Debra Loftus, Mark Puchalski, and Richard Zinbarg and current students Kathy Nugent and Eric Rosenberg, has been very influential in my thinking. As should be obvious, many of the theoretical ideas that I will discuss have been adapted from prior work of Donald Broadbent, as well as John Atkinson, Hans Eysenck, and Jeffrey Gray.

[2] It is important to note that while Broadbent and his colleagues pointed the way, a few other cognitive psychologists have not completely ignored individual differences, for there has been a some interest among experimental psychologists in individual differences in cognitive ability. Until recently, however, with the possible exception of the application of cognitive modeling techniques to the study of intelligence (e.g., Hunt, 1983), the two disciplines discussed by Cronbach (1957, 1975) and Eysenck (1966), have not moved much closer together than they were 35 years ago. The thrust of this chapter is not the study of individual differences in cognitive ability, however, but rather the study of how seemingly non-cognitive or affective differences relate to cognitive performance. Examples of recent applications of cognitive theory to the study of psychopathology (M. W. Eysenck and Mathews, 1987; MacCloud & Mathews, 1991) or to the study of mood represent a welcome resurgence of interest in viewing individual differences in affect as parameters that are important sources of variance in cognitive models.

[3] The study of stress may be divided into an examination of the effect of a variety of situational manipulations versus analyses of physiological response systems (Sanders, 1983). Broadbent and his colleagues have focused on the former; Selye (1976) and his followers on the latter.

[4] This concept of parallelism is the opposite view to those who apply additive factors logic to task decompositions. In the additive factors model, interactions are seen as suggesting effects at similar stages, additive effects as showing that different stages are affected. Parallel effects are found when two variables have similar patterns across a variety of tasks. An early example of this was the observation that extraversion had a pattern of effects that was similar to that of sleep deprivation. More recent examples include the parallel effects of lesions to the septal-hippocampal loop and the effects of anti-anxiety drugs (Gray, 1982).

[5] This approach of studying personality variables in combination with environmental manipulations is one that has been advocated by others as well. Psychometricians such as Cronbach (1957, 1975) and Snow (1989) have long argued for the need to combine theories of individual differences with theories of the situation. Although many have proclaimed the need to study interactions of personality and situations, the best work in terms of programmatic theory construction and testing has been done by Hans Eysenck and his associates (Eysenck, 1967, 1981, 1991) and by John Atkinson and his associates (Atkinson, 1957; Atkinson and Birch, 1970; Atkinson and Raynor, 1974; see also Brown & Veroff, 1986). It is a combination of the Atkinson, Eysenck and Broadbent traditions that has most influenced my own work.

[6] The "Big 5" are thought to represent the fundamental axes of personality on the grounds that humans have developed language to describe important events and individual differences in their world. Thus, it is thought to be important to describe the dominance or energy level of a conspecific, as it is the trustworthiness and emotional stability of a potential colleague or spouse. However, it is also possible that language has developed to describe important combinations of underlying biological variability. For it is very important to describe those who are sensitive to reward and not sensitive to punishment and very aggressive (psychopathic killers), and thus it is perhaps more likely to develop words to describe combinations of underlying dimensions than words to describe the endpoints of the dimensions themselves.

[7] A synthesis of the five and three factor models notices that individual differences can be categorized in terms of the valence of emotional states normally experienced by individuals and in the different valence of emotional states one's behavior produces in others. Thus Extraverts or highly Surgent individuals are more sensitive to positive mood states than Introverts or less surgent individuals (Eysenck, 1952; Larsen & Kasimatis, 1990); Neurotics and emotionally unstable individuals are more sensitive to negative mood states than are emotionally stable individuals (Eysenck, 1952; Larsen & Kasimatis,1990); Agreeableness seems to induce positive affect in others; and finally, Psychoticism or a lack of Conscientiousness induces negative affect in others.

[8] In addition, these same personality traits affect behavior at the different temporal levels described in Figure 1. A broader discussion of how personality and motivational variables affect choice, persistence and growth throughout the lifespan is clearly beyond the scope of this chapter. The stages of processing outlined in Figure 3 represent times of less than a few seconds, although effects of arousal on storage has been examined for longer periods of time..

[9] Given the elegant simplicity of the models associated with Donald Broadbent, it is problematic to present such an overly complex model. I hope that the tolerant reader will recognize that this model tries to capture the cognitive processes considered by Broadbent as well as the differential sensitivities to cues for positive and negative outcomes discussed by Gray (1981) and the general theory of emotional reactivity discussed by Rolls (1990). This figure was initially prepared to integrate the work of Clark and Watson (1991), Larsen (1991), and Thayer (1989) which has been based upon the prior theories of Broadbent (1971) and Gray (1972).