The Implications of Arousal Effects for the Study of Affect and Memory


As we suggested earlier, when relating arousal to cognitive performance an important source of variation is individual differences in tonic as well as phasic arousal. Although it is possible to ignore such individual differences by random assignment of subjects to affective manipulations, a more powerful design attempts to take these individual differences into account. Not only does so doing increase statistical power, it also leads to a clearer explication of the relationships between affect, arousal, and memory. In this section we discuss alternative strategies for identifying individual differences associated with arousal, and then review some of the most relevant prior research.

There are two very different ways to examine these individual differences: one is to search for behavioral correlates of psychophysiological variation; the other is to search for psychophysiological correlates of variations in personality traits known to be behaviorally important. Given our orientation towards developing and testing theories of personality, it is not surprising that we prefer the latter orientation.

Bottom up versus top down searches for individual differences. A typical finding in the psychophysiological literature is that although there are large individual differences on each measure, the separate measures do not tend to correlate very highly. This has been taken, in fact, as evidence of the separability of various types of arousal. That is, it is the low correlation between measures that makes one doubt the utility of a general factor of arousal. To anyone familiar with scale construction, such a pattern of results is not surprising. Individual items in an ability test show very low average correlations with other items (of the order of .1 to .2), even in a test that has a very high reliability (.95). That is, although the general factor in such a test might account for no more than 10 or 20% of the reliable variance in any particular item, 95% of the total test variance is due to the general factor. When using such a test to predict an important criterion, it is difficult to find high validity coefficients for any particular item (given the low saturation of the general factor in each item) even though the test as a whole has a strikingly high validity.

Those who search for reliable individual differences in arousal should consider this example carefully, for it suggests that the appropriate strategy of searching for useful individual differences is not in searching for high validity coefficients at the item level (separate measures of physiological response) but rather at a higher, test level (composite measures representing a greater sampling frame or a greater number of measures).

That is, an alternative strategy to looking for correlates of specific physiological measures is to reverse the search and look for physiological correlates of personality dimensions that are known to be important. As we show below, perhaps the most important of these is the dimension of introversion-extraversion and its component, impulsivity.

Dimensions of personality related to arousal. Rather than try to summarize all studies of arousal that have shown individual differences (which would be all studies with more than one subject), we have found it useful to take a top down approach and discuss how certain broad, stable dimensions of individual differences can be related to arousal. To help the uninitiated, this means considering the relevant personality literature to suggest dimensions that on theoretical grounds should be related to arousal. For a much more thorough review, see Gale and Edwards (1986).

Several important conclusions have come out of the last 50 years of personality research. One is that there are a limited number of dimensions (between four and seven) of personality that can be used to describe individual differences in cognitive, affective, and social behavior. The second is that individual differences in these trait dimensions have remarkable stability across years and even decades. A third is that these traits have important relationships to meaningful behavioral outcomes. Lastly, some of these dimensions have been shown to have consistent relationships with various physiological measures.

That there is a limited set of dimensions that can usefully describe individual differences in social behavior has been shown by both the American steps toward an "adequate taxonomy" of folk psychological descriptions of personality (e.g., Allport & Odbert, 1936; Cattell, 1957; McCrae & Costa, 1987; Digman, 1990; Goldberg, 1982; Norman, 1963, 1969) as well as the European studies on the biological bases of personality (H. J. Eysenck 1947, 1952, 1967, 1981, 1987; H. J. Eysenck and M. W. Eysenck, 1985; Gray, 1972, 1981, 1982, 1987; Strelau, 1983, 1985). The American taxonomy project of natural language or folk psychology has suggested that five dimensions of describing others (e.g., the five factors discussed by Tupes and Christal, 1961, and Norman, 1963, 1969: surgency, agreeableness, conscientiousness, emotional stability, and culture) are the beginnings of an adequate taxonomy. The British and other Western Europeans have intensively studied three-four dimensions (introversion-extraversion, neuroticism-stability, psychoticism, and intelligence) while the Eastern Europeans have concentrated on three dimensions derived from Pavlovian theory (strength of the excitory and inhibitory processes, and balance between these process, Strelau, 1983).[5]

Recent reviews of longitudinal studies involving this limited set of dimensions has shown quite high consistencies across periods of up to forty years (Conley, 1984). For example, while there are clear changes across the lifespan, measures of introversion in early adulthood predict social behavior 30-40 years later. Similarly, young adults high in neuroticism are more at risk for psychiatric difficulties over the life span than are those with low neuroticism scores.

Most important for a discussion of personality and arousal, the trait of introversion-extraversion, or at least its component, impulsivity, has a striking relationship to measures of arousal. Several physiological factors appear to differ between these groups. These include differences in basal arousal level (introverts are claimed to be more aroused than extraverts), differences in speed of habituation of arousal (extraverts and high impulsives seem to habituate faster) and differences in phase of the diurnal arousal rhythm (low impulsives seem to lead high impulsives by at least several hours). These differences have been reported using a variety of diverse measures.

In several summaries of the EEG-personality relationship, Gale (1981, 1983) has suggested that while there are reliable differences between introverts and extraverts, they are only to be found in certain, well defined circumstances. Gale has commented that the demands of experimental conditions used in studying personality--EEG correlates vary greatly, ranging from instructing subjects to lie back in a dark room and try not to fall asleep, to having them count tones of moderate intensity, to trying to learn which extremely loud tones are associated with shock. In a post hoc classification of experimental situations, Gale found evidence for Eysenck's hypothesis of greater arousal for introverts for those studies with moderate levels of stimulation.

Other cortical measures suggesting higher arousal for introverts can be found from studies using Evoked Potentials (ERP; Stelmack & Wilson, 1982), Contingent Negative Variation (CNV; O'Conner, 1980,1982; Werre, 1986, 1987), Photic Driving (Robinson, 1982) and the Sedation Threshold (Claridge, Donald, & Birchall, 1981; Claridge & Ross, 1973; Shagass & Jones, 1958; Shagass & Kerenyi, 1958).

Introversion-extraversion and impulsivity also relate to individual differences in arousal as assessed by autonomic measures of Skin Conductance (Fowles, Roberts and Nagel, 1977; Smith, Rympa and Wilson, 1981; Wilson, 1990), Spontaneous GSRs (Crider & Lunn, 1971), Orienting Response Habituation (O'Gorman, 1977; Stelmack, Bourgeois, Chian and Pickard, 1979), and Pupil Size (Stelmack & Mandelzyz, 1975); phase differences in the basic metabolic measures of diurnal variation in Body Temperature (Blake, 1967); and differences in self reported arousal (Thayer, 1989).[6]

Individual differences in arousal as measured by self report

Two dimensions of self reported arousal. Although much simpler than measuring EEG, SC, or OR, a powerful device for assessing arousal state is self report. Subjects will report being more lively, active, full of pep or energetic as a function of time of day, prior exercise, or stimulant drugs. Thayer (1967, 1978, 1989) has shown that his Activation-Deactivation Adjective Check List (AD-ACL) correlates with the appropriate physiological measures and is quite sensitive to arousal manipulations (see also Clements et al, 1976). Although Thayer originally reported four factors to his AD-ACL, later analyses suggested that two factors were quite adequate: activation-deactivation (energetic arousal) and tension-relaxation (tense arousal). The former dimension shows a marked diurnal rhythm while the later one does not. In addition, it is commonly found that the energetic arousal factor is a more reliable predictor of performance (Loftus, 1990; Matthews, Davies & Lees, 1990; Thayer, 1978).

Individual differences in diurnal variations of self-reported arousal. It is well known that some people claim they function best in the morning and need to retire early, while others believe that they can not function at all until late in the day. Individual differences in this lark versus owl tendency are reflected in body temperature differences as well as in such simple measures as time of awakening. Although in general, owls are more likely to be highly impulsive, the correlations are not particularly strong. Matthews (1989) has made use of this fact to study the performance of introverts and extraverts (and low and high impulsives) as a function of time of day and self-reported arousal. His results parallel those found with direct manipulations of arousal. That is, low impulsives who report being aroused in the morning perform less well than those who report not being aroused, while high impulsives who report being aroused do better in the morning than those who do not.

Individual differences in self-reported arousal and affect

We have emphasized that there are stable individual differences in tonic and phasic arousal not just because such differences are an extraneous source of variability when studying affect, but because the same dimensions of personality that relate to arousal relate to sensivity to positive and negative affect. Recent work has shown that positive and negative affect are independent of each other (Watson & Tellegen, 1985). In addition, it is thought that susceptibility to positive affect is related to impulsivity and extraversion and that susceptibility to negative affect is related to anxiety and neuroticism (Meyer & Shack, 1989). This suggests that most manipulations of mood or affect are likely to have different effects depending upon individual differences in impulsivity and anxiety.