The personality theories of H. J. Eysenck and J. A. Gray: a comparative review

The personality theories of H. J. Eysenck and J. A. Gray:a comparative review

Gerald Matthewsa, *, Kirby Gillilandb

aDepartment of Psychology, University of Dundee, Dundee DD1 4HN, Scotland, U.K.bDepartment of Psychology, University of Oklahoma, 455 West Lindsay, Norman,

OK 73019, U.S.A.

Received 6 November 1997

Abstract

Hans J. Eysenck and Je�rey A. Gray have proposed in¯uential theories of the biological bases ofpersonality traits. Eysenck's theory concerns the extraversion, neuroticism and psychoticism traits,whereas Gray proposes the use of new, rotated axes of impulsivity and anxiety. Eysenck usesmultiple arousal systems as the central explanatory constructs, whereas Gray describes more speci®csystems related to behavioural inhibition and activation. This article reviews the evidence relating tothese theories provided by studies of c.n.s. and a.n.s. psychophysiology, subjective a�ect,conditioning and attention and performance. It discusses key predictive successes and failures andmethodological problems which may impede theory-testing. It is concluded that there is a solid coreof predictive support for the Eysenck theory in some paradigms, such as the moderator e�ect ofstimulation level on individual di�erences in phasic electrodermal response and eyelid conditioning.In other settings, the theory fails to explain empirical data adequately, especially in studies ofsubjective response and attention and performance. Gray's theory has advanced research throughstimulating interest in moderation of personality e�ects by motivational variables. It also provides abetter explanation than Eysenck's theory for certain data, such as instrumental conditioning toreward stimuli and the positive a�ectivity of extraverts. Overall, however, Gray's theory explains anarrower range of ®ndings than Eysenck's. There is little evidence that Gray's revised personalityaxes are generally more predictive of psychophysiological and performance criteria than Eysenck'soriginal dimensions. Finally, it is suggested that the assumptions of the biological approach topersonality are in need of reassessment. It is possible that the biological theories may be improvedthrough developments in methodology or through discriminating multiple systems underpinningtraits. For example, extraversion may have distinct ``reticulo±cortical'' and ``dopaminergic'' aspects.Alternatively, the biological approach may not in fact be adequate for explaining behavioural

Personality and Individual Differences 26 (1999) 583±626

0191±8869/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reservedPII: S0191±8869(98)00158±5

PERGAMON

* To whom all correspondence should be addressed. E-mail: [email protected]

correlates of traits. In this case, trait research should place more emphasis on cognitive or socialbases for personality. # 1999 Elsevier Science Ltd. All rights reserved.

1. Introduction

Hans Eysenck and Je�rey Gray have been among the foremost exponents of the hypothesisthat personality traits provide a window on individual di�erences in brain functioning. Boththeorists start from the assumption that we can characterise brain processes by means of asimpli®ed ``conceptual nervous system'' comprising the key circuits relevant to personality andbehaviour. The multitude of associations between personality and behaviour establishedempirically may derive from individual di�erences in quite simple parameters of brain function.Gray's personality theory began as a modi®cation to the Eysenck theory, but is now usuallyseen as an alternative theory (cf. Gray, 1981). In this paper, we review data from experimentalstudies testing predictions from the two theories and conclude with some thoughts on theirrespective merits.

1.1. Conceptual nervous systems for personality

Eysenck (Eysenck, 1967; Eysenck and Eysenck, 1985) identi®es two principal brain systemsas the key components of his conceptual nervous system: reticulo±cortical and reticulo±limbiccircuits. The reticulo±cortical circuit controls the cortical arousal generated by incomingstimuli, whereas the reticulo±limbic circuit controls response to emotional stimuli. Understrong emotional stimulation, limbic system activity may spread to the cortex. Extraversion±introversion (E) relates to arousability of the reticulo±cortical circuit, so that introverts aretypically more aroused than extraverts. However, methodological analyses of extraversionstudies (notably Gale, 1973) have illuminated two basic problems for theory testing. First,people actively seek a moderate level of arousal, so that relationships between personality andarousal may also re¯ect individual di�erences in strategies for seeking or avoiding stimulation.For example, in sensory deprivation studies, extraverts seem to raise their arousal by movingaround (Eysenck and Eysenck, 1985) and extraverts generally prefer more arousing activities(Furnham, 1981).Second, under high levels of stimulation, a protective ``transmarginal inhibition'' (TMI) may

lead to paradoxically reduced arousal. According to Eysenck (1994), increasing stimulationprovokes increasing c.n.s. reactivity until an optimal point is reached, beyond which inhibitionand decreasing reactivity set in. Eysenck (1994, p. 161) states that `Ìn the case of the mostarousing testing conditions we might suspect that the optimal point for introverts, but not forextraverts, had already been passed, so that extraverts' arousal was still growing, introverts'declining''. Hence, introverts may be higher, lower or equal to the arousal level of extravertsaccording to complex interactions of personality type and environmental manipulation.Careful, theory-driven control of experimental parameters is thus essential (H. J. Eysenck,1981).

G. Matthews, K. Gilliland / Personality and Individual Di�erences 26 (1999) 583±626584

Neuroticism-stability (N) is associated with arousability of the limbic circuit, such thatneurotics become more aroused than stable individuals as a consequence of emotion-inducingstimulation. Hence, individual di�erences in N may only be evident in emotional or stressfulcontexts. The neuropsychology of Eysenck's third dimension of Psychoticism (P) has not beenworked out in detail. H. J. Eysenck (1992) suggests that P is inversely related to serotonergicfunction, but, more recently (Eysenck, 1997), P is linked to dopamine.

Gray's theory modi®es both the psychometric alignment of the major personality dimensionsand, to some extent, their biological bases. Originally, Gray (1970) proposed two newdimensions of anxiety (Anx) and impulsivity (Imp) oriented at 308 to Eysenck's E and N,respectively. Gray's model is often assumed to posit a 458 rotation, but Gray (personalcommunication, 18/11/97) points out that this is a misconception derived from the way themodel was presented graphically (e.g. Gray, 1981). Here, we shall assume a ``308 rotationmodel'', which aligns Anx most closely with N and Imp most closely with E. Anx alsocorrelates to some degree with introversion and with low P, so that Anx may be contrastedwith risk-taking ``primary'' psychopathy (Gray, 1987a). A second, orthogonal Imp dimensioncorrelates most highly with E, but also correlates to some extent with N and P. Gray's (1987a)conceptual nervous system is shown in Fig. 1. Anxiety is associated with sensitivity of one ofthese components, the behavioural inhibition system (BIS), which is activated by fear andnovelty stimuli and signals of punishment and non-reward. The core of the system is a septo-hippocampal comparator which detects mismatch between the actual and predicted state of theworld. Mismatch detection results in inhibition of ongoing behaviour and increased arousaland attention. The comparator works in conjunction with other brain structures, such as theprefrontal cortex, permitting some cognitive control over anxiety in humans and variousascending a�erents from more primitive levels of the brain. Gray (1987a) discusses the neuralsystem supporting Imp, the behavioural activation system (BAS), more brie¯y. It is sensitive tosignals of reward and non-punishment and in¯uences probability of approach behaviour. Gray(1987a, 1991) sees the BAS as associated with mesolimbic dopaminergic pathways ascending

Fig. 1. A conceptual nervous system for personality (adapted from Gray, 1987a).

G. Matthews, K. Gilliland / Personality and Individual Di�erences 26 (1999) 583±626 585

from nucleus A10 of the ventral tegmentum of the brainstem (see also Gray et al., 1991). Grayagrees with Eysenck that a third major dimension of personality is required, aligned at leastapproximately with psychoticism, which Gray (1991) relates to the ®ght/¯ight system.

1.2. Levels of description within the theories

The two theories di�er in the level of physiological construct preferred, as well as in thespeci®c circuits linked to personality. Eysenck's reticulo±cortical and limbic systems arebroadly de®ned and system components such as the ascending reticular activating system(ARAS) may be comprised of a number of structures and pathways. Gray (1987a) identi®esarousal as a system separate from BIS, BAS and ®ght/¯ight, although these systems are indynamic interaction. However, Gray (1987a) uses `àrousal'' in a more restricted sense thanEysenck, to refer to the dorsal noradrenergic bundle (DNAB), ®bres which ascend from thelocus coeruleus to innervate many forebrain structures. Although some researchers (e.g.Panksepp, 1982) relate noradrenergic activity to a general emotional arousal function, theDNAB does not seem to constitute the generalised arousal system envisaged by the pioneers ofarousal theory. Robbins (1986) suggests that the DNAB may relate to a compensatorymechanism which maintains e�ciency of selective attention and other functions during statesof stress. Hence, some functions traditionally related to arousal may not be attributable toGray's arousal system, including (1) motor responsiveness, which perhaps relates most closelyto dopaminergic pathways (cf. Gray et al., 1991) and (2) sensory and attentional functionsrelated to central cholinergic a�erents to the cortex (Beatty, 1986), whose place in Gray'sconceptual nervous system is unclear.One of the strengths of Eysenck's theory is that the use of arousal as a mediating variable

allows personality to be linked to many qualitatively di�erent response indices. Gray's theorytrades o� this generality of application for more ®ne-grained neuropsychological descriptionand greater speci®city of prediction of a narrower range of behaviours. Often, however, it isdi�cult to establish correspondences between human behavioural responses and the outputs ofGray's systems. In such cases, Eysenck's arousal theory may provide a basis for predictionwhere Gray's more neurologically detailed theory is silent. Hence, in contrasting the Eysenckand Gray theories we are not altogether comparing like with like. Clearly con¯ictingpredictions may sometimes be drawn from the two theories, but in other circumstances the twotheories may be referring to the same underlying physiology at di�erent levels of description,especially where the slippery construct of arousal is concerned.

1.3. Criteria for theory testing: scope of the review

Research on the two personality theories has adopted several strategies for theory testing,including psychometric analyses, the comparison of normal and clinical populations andexperimental tests of predictions derived from neuropsychological theory. Eysenck (e.g.Eysenck, 1994) has also emphasised personality correlates of clinical conditions, but thetheories do not seem to generate clearly contrasting predictions concerning the relationshipbetween personality and disorder, especially as Eysenck (1997) and Gray et al. (1991) agree on


a dopaminergic mechanism for psychosis. In this article, we assess the theories against thefollowing criteria, each of which has its own advantages and disadvantages.

1.3.1. Comparison of dimensional modelsGray (1981, pp. 247±252) highlights studies which investigated which sets of personality axes

correlate best with psychophysiological and behavioural criteria. He suggested that Imp ismore predictive than sociability (Soc) of various criteria. Hence, a rough and ready way todistinguish between the theories is in terms of the predictive power of the trait variables theyspecify. If we ®nd that Imp and Anx are more strongly and coherently related to dependentmeasures, then we have an argument in favour of the Gray theory. In essence, we are seeking atype of ``simple structure'': the set of traits which align most closely with sets of responsemeasures. Given the lack of comparative work on N and Anx, most of the evidence isconcerned with comparisons between E, Imp and Soc. In recent years, several researchers haveattempted to assess the Gray dimensions using new scales (e.g. Wilson et al., 1989;MacAndrew and Steele, 1991) or by reinterpreting scales developed for other purposes.However, there have been insu�cient psychometric and experimental studies for any of thesenewer measures to be convincingly validated as measures of BIS and BAS function.

1.3.2. Comparison of moderating factorsThe theories di�er not just in their choice of dimensional axes, but also in the moderating

factors which they identify as critical. A further indication of theory validity is provided bytesting whether personality e�ects are primarily moderated by level of stimulation or arousal(as Eysenck predicts) or by reinforcement signals (as Gray predicts). The di�culty in applyingthis criterion is that the arousing and motivating properties of experimental manipulations maybe mutually confounded.

1.3.3. Direct tests of theoryComparison of dimensions and moderating factors is a somewhat indirect approach to

theory evaluation. A stronger criterion is the outcome of tests of con¯icting predictions derivedfrom the two theories within speci®c experimental paradigms. However, because of themethodological and conceptual di�culties previously described, it is di�cult to ®nd suitableindices of the underlying conceptual neural systems. Additional hypotheses are required to linkneuropsychology to overt psychophysiological and behavioural response, such as the Yerkes±Dodson Law in the case of studies of arousal and performance. Discon®rmation of predictionmay re¯ect failure of the additional hypothesis, rather than failure of the theory (see H. J.Eysenck, 1981).The present article is mainly concerned with experimental studies which a�ord direct tests of

theory. We will be concerned primarily with studies of E and N on the one hand and Anx andImp on the other, because the two theories have a similar view of P. This review is structuredaround a number of di�erent topic areas, within which the two theories make di�eringpredictions. For each area, we will assess the validity of the two theories by comparing howsuccessfully their predictions match empirical data. We will discuss whether the predictivefailures may be attributed to failures of additional hypotheses linking responses to neuralsystems, rather than to the personality theory itself. We begin with psychophysiological


correlates of personality, which we divide into central and peripheral nervous system studies.Next, we consider studies of the specialised branch of psychophysiology which deals withemotional and other subjective states. Two areas of behavioural research are reviewed:conditioning and learning and attention and performance. The behaviours concerned becomeprogressively more distant from the neural substrates described by Eysenck and Gray and soprovide a progressively greater challenge for neuropsychological theory.

2. Psychophysiology of the central nervous system

2.1. Psychophysiology of personality: theoretical and methodological issues

Owing to the biological nature of both Eysenck and Gray's theories, psychophysiologicalmeasures would appear to o�er the most direct and e�cient method for investigation.However, there are various di�culties in making and testing predictions. Gale's (e.g. 1973,1983) series of methodological reviews of the EEG refers to problems with subject selection,bias within the subject or experimenter, a lack of theoretical sophistication, over-attention tooutcome at the expense of process, poor psychometrics or psychophysiology, trivialexperimentation and procedural insensitivity (see also critical reviews by O'Gorman, 1984;Eysenck, 1994). Four fundamental di�culties are as follows:

1. It is hard to obtain an accurate translation from theory-relevant elements of theneurological substrate to the level of discrete psychophysiological measures.Psychophysiology is not neurophysiology (Gale, 1973; Zuckerman, 1991).

2. Gale (1973) identi®es a multitude of confounding variables that in¯uence arousal and theintegrity of speci®c psychophysiological measures, so that careful experimental control ofthe experimental context is essential.

3. The validity of speci®c psychophysiological responses as arousal indices is threatened byphenomena such as directional fractionation and response stereotypy (Lacey, 1967).

4. Understanding of both personality traits and the neuroanatomical substrates ofpsychophysiological indices changes over time as constructs are clari®ed, but researchers donot always make methodologically and theoretically informed selections of personality scalesand psychophysiological indices.

These di�culties may leave the impression that the enterprise of seeking relationshipsbetween personality and psychophysiology is much like aiming at a moving target withwobbling sights. Nevertheless, several commentators consider that robust, theoreticallysigni®cant relationships between personality and psychophysiological responses may beobtained using suitable methods (see Geen, 1983; Eysenck and Eysenck, 1985; Stelmack, 1990;Eysenck, 1994 for reviews). More recent studies seem to be more rigorous than those criticizedin Gale's reviews, in general, although some continue to ®nd shortcomings (Zuckerman, 1991).In this article, we focus on the more programmatic, theory-driven and methodologicallyrigorous studies.


We have organized the studies reviewed into two broad categories: those using centralnervous system (c.n.s.) measures and those using autonomic nervous system (a.n.s) measures.These categories are both convenient and theoretically relevant. (We have omitted discussion ofbiochemical and brain-scanning studies as being too far removed from the main theoreticalissues.) As discussed above, the Eysenck theory predicts that both E and N should relate toc.n.s. and a.n.s. arousal. In the case of E, relationships should be moderated by level ofstimulation and in the case of N by levels of ``stress'' or emotion.Prediction from the Gray theory is more di�cult. If both BIS and BAS activity feed into

greater noradrenergic arousal, we expect impulsives (and extraverts) to show more arousal inresponse to reward signals, whereas anxious individuals (and neurotics) should be moreresponsive to punishment cues. However, it is somewhat unclear which arousal responses aresensitive to DNAB activity (as opposed to other arousal systems). Furthermore, there may bepsychophysiological responses which are directly driven by the BIS or BAS, withoutinvolvement of the separate arousal system, although, again, Gray does not specify theseresponses in detail. Fowles (1980) has argued that electrodermal measures are probably a goodindex of the BIS and that heart rate is a good measure of the BAS.

2.2. The electroencephalogram

There are numerous psychophysiological measures of c.n.s. activity. Perhaps the most wellknown is the electroencephalogram (EEG), the measure of the raw electrical activity producedby the brain, derived from passive electrodes placed on the scalp. The complex EEG waveformis often analyzed by decomposing the waveform through spectrum analysis into various rangessuch as delta (<4 cps), theta (4±8 cps), alpha (8±13 cps) and beta (>13 cps). The alphafrequency range has been classically associated with low states of arousal or relaxation, whilethe beta range is most often associated with wakeful, active states.Eysenck (1994) has referred to the EEG, especially the alpha range, as the ``standard

measure of cortical arousal, ever since the discovery of the ARAS'' (p. 167). Studies of therelationship between raw EEG measures and E have been reviewed almost exhaustively (e.g.Gale, 1983; O'Gorman, 1984; Zuckerman, 1991; Eysenck, 1994). While it is unnecessary torecount this body of literature yet again, some attention to the reviews and their conclusions ispro®table.The earliest of the systematic reviews was a landmark analysis by Gale (1973). According to

Gale, several studies supported the hypothesis that introverts are higher in cortical arousalthan extraverts. However, a similar number found no di�erences and three studies foundresults that contradicted the theory. Gale (1973) observed that the arousal inducing propertiesof the testing environment appeared to mediate the ®ndings of the studies reviewed.Gale argued that moderate arousal-inducing environments, including opening and shutting theeyes upon instruction, were the most amenable to testing predictions of Eysenck's (1967)theory. Low arousal-inducing environments, such as when the subject was simply lying witheyes closed, were believed to result in paradoxical arousal, especially in extraverts and higharousal-inducing environments involving task performance demands were believed to result inpossible over-arousal, again especially for extraverts. After post hoc reordering of the studiesaccording to the arousal inducing properties of the testing environment, Gale revealed that in


studies incorporating moderate arousal levels, introverts exhibited greater evidence of alphaactivity re¯ecting lower levels of arousal, as compared to extraverts, as predicted. A laterreview by Gale (1983) developed both the methodological critique and reviewed results from anew decade of con¯icting EEG research on E, arriving at similar conclusions to the earlierreview.O'Gorman (1984) contended that Gale may have made some errors of omission and

classi®cation. His own, alternative analysis of the EEG studies found that use of Eysenck'soriginal E scales or similar measures, was a more important in¯uence on outcome than arousallevel (see Gale, 1984, for a rejoinder). Zuckerman (1991) has been less than enthusiastic aboutthe level of support that past EEG studies have provided for the cortical arousal hypothesis ofextraversion, although he points out that studies using female subjects or equal numbers ofboth sexes seem to more often support Eysenck's theory.There have been a few more recent studies that are noteworthy. In a sample of 180 subjects,

Matthews and Amelang (1993) measured power values in three bands in three experimentalconditions, whose levels of stimulation corresponded to Gale's three environment types.Signi®cant correlations between personality and EEG measures averaged across environmentswere low in magnitude, not exceeding 0.20, but broadly matched expectation. Extraverts andimpulsives showed more slow-wave activity (but not more alpha) and neurotics showed morebeta activity. Analyses of the individual environments showed little systematic di�erence inpersonality correlates across di�erent levels of stimulation. In the ``moderate arousal''condition considered by Gale most likely to support the Eysenck theory, both E and Imp wereunrelated to alpha, but positively correlated with beta. Matthews and Amelang suggest thatpersonality traits are only weakly predictive of the spontaneous EEG and most publishedstudies have far too few subjects for assessment of potential moderators of personality±EEGassociations.Smith et al. (1995) reported that introverts were generally found to produce lower levels of

alpha activity re¯ecting higher levels of arousal, but there were also complex hemisphere bygender interactions. Two groups showed asymmetry of alpha: both female introverts andextraverted males exhibited higher alpha activity on the right side. These ®ndings werecomplicated by the fact that subjects were selected to be high in N because the study focusedon emotional reactivity. Aside from providing, once again, some conditional support for thearousal hypothesis of E, these ®ndings suggested that concerns about sex of subjects may beimportant (Zuckerman, 1991), as well as electrode placement (Gale, 1983).Finally, Stenberg (1992) attempted to contrast potential di�erences between the E and N

dimension of Eysenck and the Imp and Anx dimensions of Gray in a study which manipulatedpositive and negative emotional imagery. Subjects were classi®ed according to the EPI and theKarolinska Scales of Personality. Factor analysis of the multiple EEG sites used revealed twofactors associated with arousal and three factors that were associated with responses toemotional reactivity. The EPI E and N dimensions were not signi®cantly related to any of theEEG factors. However, factor analysis of the joint personality scales revealed factors thatresembled Gray's Imp and Anx dimensions. More impulsive subjects showed signs of lowerEEG arousal than low impulsive subjects and more anxious subjects showed greater right-sidefrontal theta activity across all conditions, suggesting higher emotionality. Furthermore, highanxious subjects evidenced higher beta rhythm activation to the negative emotional condition,


but the high impulsive subjects did not show a corresponding reaction to the positiveemotional condition. These results suggest use of Gray's dimensions is preferable, but it isunclear to what degree the derived dimensions actually represented Gray's axes and the samplesize (40) was insu�cient for factor analysis. Stenberg concludes that there is much moresupport for the view that Imp is associated with low arousal than with facilitation of positivea�ect.So what can be said of the EEG studies of E? Gale's (1973) original critique has led to a

well articulated debate on the many sources of experimental error in this area. Nevertheless,despite the con¯icting results, methodological problems and thoughtful and well-reasonedarguments to the contrary (O'Gorman, 1984; Zuckerman, 1991), there is support for anarousal hypothesis for E. Reviewing e�ects of N and Anx, Zuckerman (1991) concludes thatthe precise nature of the relationships between these traits and the EEG is unclear. Signi®cante�ects of N are not uncommon but their direction varies from study to study and N frequentlyinteracts with E. There is some evidence (O'Gorman and Lloyd, 1987) that Imp may beespecially predictive of the EEG, in line with Gray's theory. However, in the absence ofmanipulation of motivational signals, the data have limited implications for the theory. A newgeneration of investigations that di�erentiate between the Gray and Eysenck theories isrequired, in which the stimulating and reinforcing properties of experimental manipulations areclearly distinguished.

2.3. Event-related potentials

Another index of c.n.s. activity is the event-related potential (ERP) or evoked response,derived from averaging EEG samples across trials (see Coles et al., 1990). The majority of ERPstudies have investigated responses that occur in the ®rst 100±500 ms following a stimulus,although time periods as short as 10 ms are associated with the auditory brainstem response.The major component waveforms are usually related to cognitive processing as follows. Earlycomponents (N100, P200) relate to sensory properties of stimuli and to selective attentionprocess. Later components are more ``cognitive''. The P300 has been associated with processesrelated to classifying or updating memory representations of stimuli. The amplitude of theP300 increases as the signi®cance of the event and its relevance to the subject increases and asthe demand for cognitive resources increases. The latency of the P300 measure has been relatedto the time needed to categorize and evaluate the stimulus, which appears to be independent ofthe time needed for response-related processes. Most researchers using the P300 see it as anindex of cognitive rather than arousal processes (Coles et al., 1990), but, as Eysenck (1994) hasnoted, ``The fact that the activities associated with the P300 (habituation, orienting responses,stimulus classi®cation) are closely related to the concept of cortical arousal suggests that testsof the introversion-arousal hypothesis could use the P300 paradigm with advantage'' (p. 172).Reviews of ERP measures and extraversion (Geen, 1983; Zuckerman, 1991; Eysenck, 1994;

Stelmack and Houlihan, 1995) reveal the same degree of variation in subjects, testingconditions, stimulus characteristics and other variables as those seen in the EEG studies. Earlystudies of ERP measures were also notable for their inconsistent results, probably owing tonumerous experimental control factors. More recent studies show a trend for amplitudes ofboth the earlier somatosensory components and the later P300 to be higher in introverts than


extraverts. The ®nding of higher P300 amplitudes in introverts has typically been interpreted assuggesting that introverts are experiencing greater levels of attentional demand (Stelmack andHoulihan, 1995), a result which may, in turn, be suggestive of higher arousal in introverts(Eysenck, 1994). However, failures to replicate e�ects even across similar studies is a cause forconcern (Zuckerman, 1991). In addition, the increased P300 amplitude seen in introverts maynot be evident in initial trials or sessions with small numbers of trials. Rather, this di�erencemay be the result of greater habituation in extraverts (Di Traglia and Polich, 1991) and thusmore likely to occur in sessions with a larger number of trials or in later trial blocks.The modal ®nding is for larger P300 amplitude in introverts, but a few studies show the

reverse relationship. Stenberg's (1994) review of these ®ndings suggests that P300 is larger inextraverts when the task is of brief duration or the task is complex and cognitively demanding.Brocke et al. (1997) varied level of stimulation experimentally and showed that the negativeassociation between E and P300 amplitude became positive when subjects were exposed to60 dB noise. This e�ect might be attributed to TMI, which is initiated at lower levels ofstimulation in introverts (H. J. Eysenck, 1981). There is little evidence for any consistentdi�erence between extraverts and introverts in P300 latency, but faster P300 latency has beenreported for high N individuals (Stelmack and Houlihan, 1995).Few ERP studies have attempted di�erential tests of Eysenck and Gray's theories. However,

Bartussek and his colleagues reported three investigations that were speci®cally designed to testGray's theory. In one study, ERPs were recorded from introverts and extraverts to neutraltones and tones signifying gains or losses in a betting task (Bartussek et al., 1990a). In asecond study, ERPs were recorded to emotionally positive, negative or neutral adjectives(Bartussek et al., 1990b), so as to di�erentiate between Gray's BAS (reward seeking) and BIS(punishment avoidance) components. In both studies, introverts had greater amplitude of theERP in the neutral condition and extraverts had greater amplitude of the ERP in both thereward/positive and loss/negative conditions. These results failed to support Gray's theory andcould be viewed as supporting Eysenck's more general arousal model, provided emotionalstimuli induce TMI in introverts. In a third study, auditory ERPs were recorded to stimuli thatsignalled winning or losing in a gambling task (Bartussek et al., 1993). A statistically signi®cantpersonality by condition (win/lose) interaction was found for the ERP P200 component thatconformed to the predictions of Gray's theory, i.e., extraverts had higher amplitude in the wincondition and about equal amplitude in the loss condition, as compared to introverts. Therewas a complex interactive e�ect of E, N and condition on P300 amplitude, such that data wereconsistent with Gray's theory only in high N subjects. However, De Pascalis et al. (1996)conducted a similar study and failed to ®nd any e�ect of E, although they found somerelationships between scales of the Gray±Wilson questionnaire (Wilson et al., 1989) and somewave components. The study also showed that high N subjects showed larger N800 amplitudesto punishment stimuli, a result which seems compatible with Gray's theory, although, as theauthors point out, the functional signi®cance of N800 is unclear.Finally, Bartussek et al. (1996) tested for interactions between E and motivational stimuli in

two studies, one requiring processing of words of di�ering valence and one using a startleresponse paradigm. In both studies complex interactions involving E, stimulus valence,electrode site and other factors were found, for P300 in the ®rst study and P200 in the second.E�ects bore little resemblance to e�ects predicted from the Gray theory. Bartussek et al. (1996)


suggest that extraverts tend to develop stronger arousal at frontal sites in response toemotional stimuli, irrespective of valence. Such an e�ect might be seen as a further instance ofthe TMI predicted by the Eysenck theory. However, the Bartussek studies provide littleevidence for enhanced response to emotional stimuli in high N subjects, although Nparticipated in various complex interactions. According to the Eysenck theory, N rather than Eshould be the main factor moderating response to emotional stimuli. While any de®nitiveconclusion based on these studies remains elusive, they do represent experimental e�orts ofprecisely the type needed if we are to advance an understanding of either or both theories.Another ERP method that has received increasing attention is the auditory brainstem evoked

response (ABR: see Jewett et al., 1970 and Bullock and Gilliland, 1993 for reviews). The ABRproduces a remarkably stable and replicable series of seven waves that occur in the ®rst 10 msfollowing an auditory stimulus. Its various wave peaks relate to speci®c neural generators atsuccessively higher levels of the auditory pathway, ranging from the acoustic nerve (waves Iand II) to initial cortical projection activity (wave VII). Faster peak latencies or interpeaklatencies (indexing conduction times) re¯ect higher levels of neural activity. For many years,reports of the stability of ABRs across subjective arousal states and environmentalmanipulations of arousal suggested that the ABR was better suited to exploring peripheralrather than central nervous system processes (Eysenck, 1994; Stelmack, 1990). However, thepotential of the ABR for also exploring midbrain c.n.s. processes (Bullock and Gilliland, 1993)is shown by its sensitivity to changes in cognitive demand (e.g. Lukas, 1980) and to certaindrug manipulations that act on c.n.s. arousal centers (e.g. Church and Shucard, 1987). Thus,the ABR may be useful in re¯ecting the in¯uence of di�erential ARAS arousal or perhapsindividual di�erences in more general ARAS arousal levels, even though it is less sensitive toarousal states than other psychophysiological measures of c.n.s. activity. If reliable di�erencesin ABR activity between introverts and extraverts can be demonstrated, these di�erences mayprovide additional evidence for the physiological basis of E.Bullock and Gilliland (1993) brie¯y reviewed early studies of ABR activity and E and

concluded that the most consistent ®nding was that introverts appeared to have faster wave Vlatencies and faster wave I±V conduction times than extraverts. Bullock and Gilliland (1993)reported a study in which ABRs were recorded for introverts and extraverts during placeboand two levels of ca�eine administration, as well as during resting and two levels of responsetime task di�culty. This study included an analysis of many control factors that may havea�ected previous investigations. Introverts had faster wave V latencies than extraverts, as wellas faster wave I±III and I±V conduction times. Stelmack et al. (1993a,b) found no di�erencesbetween introverts and extraverts for any ABR wave component during sleep or wakefulness.However, the one consistent ®nding, although nonsigni®cant, was that wave V was faster forintroverts, as compared to extraverts, for all conditions. Wave V relates to inferior colliculusactivity, which is one of the auditory pathway nuclei that adjoins the ARAS. In a study acrossthe full range of E scale scores, Swickert (1996) replicated the signi®cant positive correlationbetween E and ABR wave V latency. Among the two subscales of E, Soc was found tocorrelate positively with Wave V latency, while Imp was not. This ®nding appears inconsistentwith Gray's view that Imp is the primary causal component of E, although Gray has notexplicitly addressed ABR studies. Finally, one study (Stelmack and Wilson, 1982) reportedfaster wave I latency, which led the authors to suggest peripheral neural mechanisms as the


possible site of extravert±introvert di�erences in the ABR. To the extent that wave V latencymay be interpreted as a c.n.s. measure, the results of more recent ABR studies generallysupport Eysenck's theory Ð evidence that is particularly interesting because the ABR may beanatomically consistent with locations and neural mechanisms which Eysenck relates to E.There have also been a few studies of contingent negative variation (CNV), an EEG

response emitted in anticipation of a motor response. To measure CNV researchers typicallyemploy a response time procedure that includes both a warning stimulus followed by animperative stimulus to which the subject responds. During the brief interval between thewarning and imperative stimuli, there is a negative shift in the baseline of EEG activity thatincludes an orienting component (O) and an expectancy or readiness for action (E) component(Tecce, 1972). The CNV has been viewed as an indicator of attention or arousal functions(Tecce, 1972), which o�ers the logical bridge for its use in explorations of E. In general, as theEysenck theory predicts, extraverts do have higher CNV amplitude as compared to introverts(Werre, 1986). Varying environmental and intrinsic conditions can a�ect this relationship withoptimal results being provided by subjects who were young, well motivated, isolated fromextraneous stimulation and were performing a novel reaction task in the morning (Werre,1986). Because of its links to attentional processes, the CNV has probably provided evidencefor the arousal hypothesis of E more like that from behavioral measures of attention andarousal, in contrast to raw EEG measures that are generally viewed as having more direct linksto c.n.s. processes.

3. Psychophysiology of the peripheral nervous system

At a functional level, a considerable amount of peripheral nervous system activity isregulated by c.n.s. processes. Within the peripheral nervous system, the a.n.s. is a highlycomplex and interactive neural network that, at a very simple level, is organized as twotypically opposing subsystems, the sympathetic and parasympathetic nervous systems. Crudely,the sympathetic nervous system responses assist the organism in meeting the demands ofarousal-inducing situations whereas the parasympathetic nervous system acts to conservebodily energy. Many personality researchers searching for methods of measuring bodilyfunction that re¯ect arousal or arousability have focused on psychophysiological measures ofsympathetic or parasympathetic nervous system activity. The large majority of these studieshave included electrodermal or cardiovascular measures, while a far smaller number haveincluded an array of other measures.

3.1. Electrodermal activity

Electrodermal measures attempt to quantify the electrical activity of the skin. Typically, themeasurement of electrodermal activity (EDA) involves the application of a small amount ofelectric current to the skin surface through one electrode, allowing measurement of skinresistance level (SRL) between the electrodes (or its inverse, skin conductance level; SCL). SRLor SCL are measures of baseline or tonic activity. Phasic EDA measures in response to aknown stimulus are referred to as skin resistance responses (SRRs) or skin conductance


responses (SCRs). There are many other EDA measures as well and each type of measure canhave several response characteristics, such as latency, amplitude, rise time, half recovery timeand habituation rate. It has been observed that electrodermal measures provide ``direct andundiluted representation of sympathetic activity'' (Dawson et al., 1990, p. 310), which isrelatively free from other somatic in¯uence such as heart and respiration rate. EDA measureshave traditionally been viewed as a component of the orienting re¯ex. Habituation of theorienting re¯ex has been a prime measure of inhibitory neural processes. Thus, measures ofEDA are theoretically relevant to personality research.EDA has been investigated with respect to both E and N. Eysenck's theory makes the clear-

cut prediction that EDA should be higher in neurotics, especially under stress, but moreconsistent ®ndings have been obtained with E. Electrodermal responses tend to be insensitiveto both N and trait anxiety (Zuckerman, 1991). Several reviews of EDA research on E havebeen conducted in the past several years (e.g. O'Gorman, 1977; Stelmack, 1981; Geen, 1983;Smith, 1983; Zuckerman, 1991). The general view is that EDA studies of tonic arousaldi�erences between introverts and extraverts are quite inconsistent. However, in an interestingnaturalistic study, Wilson (1990) showed that introverts had higher SCLs than extravertsthroughout much of the day, but only after age di�erences were statistically controlled. Thefrequency of spontaneous EDA has also been viewed as a measure of tonic arousal. There isan approximately equal number of studies that report no di�erences in spontaneous EDAactivity between introverts and extraverts and studies that report higher levels of spontaneousactivity in introverts than in extraverts (Smith, 1983; Stelmack, 1990).Studies of phasic EDA measures and E have shown more consistent trends. These studies

take many methodological forms, but most incorporate EDA measurement before, during andafter a test stimulus is presented (often auditory tones), sometimes in habituation ordishabituation paradigms. Stimulus and environmental characteristics play an important role inmediating the outcome of these studies (see Stelmack, 1981, 1990). For example, manyinvestigations show that introverts produce larger amplitude SCRs than extraverts whenmoderate intensity auditory or visual stimuli are used. Personality has little e�ect in studiesusing low level stimulation and at high levels of stimulation extraverts demonstrate higher SCRamplitude, perhaps due to TMI (Stelmack, 1981, 1990). Similar moderating e�ects are found instudies that have manipulated arousal level with ca�eine (see Smith, 1983), stimulus intensity(Wigglesworth and Smith, 1976) and stress (Fowles et al., 1977). These studies uniformlysupport higher levels of EDA in introverts as compared to extraverts at low levels of arousal,but not when arousal level is high.Habituation processes provide data that are less consistent. Early studies suggested faster

habituation in extraverts (O'Gorman, 1977), but more recent reviews concluded that theevidence for di�erences in trials-to-criterion during habituation between the personality groupswas simply not consistent enough to form a clear consensus (Geen, 1983; Smith, 1983). Posthabituation processes, especially dishabituation, seemed to yield fairly consistent ®ndings thatsuggest introverts are more highly aroused than extraverts (Geen, 1983).In summary, EDA measures of tonic arousal level have been either inconclusive or so

inconsistent as to add little to our understanding of arousal di�erences. Studies using phasicmeasures and tonic measures assessed during arousal manipulations have been far moreconsistent and, in general, support Eysenck's arousal model of E. Studies in this area have not


been designed to discriminate the Eysenck and Gray theories, but Fowles (1980) has linked theBIS to EDA. Unfortunately, there is general agreement that the EDA research in the area ofneuroticism has been `ùniversally negative'' (Zuckerman, 1991, p. 261; see also Fahrenberg,1987; Naveteur and Freixa-i-Baque, 1987). Furthermore, trait anxiety and N do not appear topredict either electrodermal orienting or habituation consistently, although state anxiety tendsto block habituation and relates to increased EDA (O'Gorman, 1977; Zuckerman, 1991).Contrary to the Gray±Fowles prediction, E appears to be a better predictor of EDA than N is.

3.2. Cardiovascular and other autonomic measures

Measures of cardiovascular activity have also been used to explore arousal di�erencesbetween introverts and extraverts, although, like many other autonomic measures, they are notentirely under sympathetic control. For example, the cardiovascular system interacts highlywith the respiratory system. This requires greater skill in experimental design and the controlof confounding variables that might in¯uence these measures. Most studies used heart rate as adependent measure. In the majority of cases, no e�ect of E on baseline or tonic heart rate wasreported (e.g. Myrtek, 1984; Pearson and Freeman, 1991). However, heart rate changes inresponse to stimuli or environmental changes have yielded more consistency. Greater increases(or less deceleration) in heart rate or heart rate variability for introverts, as compared toextraverts, have been reported during exposure to a series of tones (Orlebeke and Feij, 1979)and during performance of a variety of tasks (e.g. Richards and Eves, 1991). There is littleevidence for elevated heart rate in subjects high in trait anxiety or N (e.g. Naveteur and Roy,1990).Fowles (1980) has suggested that `Ìf there is to be a relationship between HR [heart rate]

and one of the three arousal systems [of Gray] ...it must be with the BAS'' (p. 91). This, ofcourse, leads to the prediction that higher levels of Imp (and thus E) would exhibit higherlevels of cardiovascular activity, a prediction quite in opposition to Eysenck's predictions andgenerally unsupported by the literature on cardiovascular activity that was reviewed above. Aswith much of the EDA research, few studies using cardiovascular measures have focused onGray's theory. De Pascalis et al. (1996) proposed that heart rate deceleration to feedbacksignals might be taken as an index of BIS activity. Neurotic subjects showed greaterdeceleration irrespective of feedback, but not the predicted di�erential sensitivity to feedback.There was no interaction between E and N, but, in addition, extraverts were more sensitive toreward signals, whereas introverts were more sensitive to punishment signals. These dataappear broadly consistent with the Gray theory, but it is curious that introverts but notneurotics should be especially sensitive to punishment signals. This study indicates a need forfurther investigation, but, overall, cardiovascular research leans somewhat in favor ofEysenck's theory.Finally, there should be brief mention of several other peripheral nervous system measures.

Support for the Eysenck arousal theory has been obtained in studies of regional cerebral blood¯ow (e.g. Stenberg et al., 1990) and pupillary response (Stelmack, 1981). Pivik et al. (1988),however, showed decreased spinal motoneural recovery in extraverts, which may be attributedto increased dopaminergic activity. This association seems consistent with Gray's theory.Mangan and Hookway (1988) showed that exposure to aversive ®lm clips induced higher heart


rate and skin conductance in extraverts than in introverts, but changes in muscle tensionshowed the opposite pattern of e�ect. There were no e�ects of N or P. Finally, there is agrowing literature on the eyeblink component of the startle re¯ex, measuredelectromyographically (e.g. Corr et al., 1997; Kumari et al., 1996). Results seem to vary withstimulus materials. Studies using unpleasant slides show enhanced response in anxiousindividuals, a result compatible with both theories. Responses to pleasant slides (which mightdi�erentiate the theories) appear inconsistent across studies (Corr et al., 1997).

3.3. Psychophysiology: conclusions

The following inferences may tentatively be drawn from the plethora of studies. First,associations between extraversion and tonic measures of c.n.s. and a.n.s. arousal are weakerthan might be expected from the Eysenck theory, although the overall trend is in the directionpredicted. Second, studies of phasic measures, including ERPs, SCRs and phasic cardiacresponse, provide stronger support for greater arousability in introverts, consistent with thetheory. Third, some of the most consistent ®ndings concern the dependence of associationsbetween E and psychophysiological response on level of stimulation, as shown most clearly inSmith's EDA work. Under high levels of stimulation, extraverts may actually appear morearoused than introverts. These results are consistent with the Eysenck theory, but require theadditional assumption of TMI in introverts. Gale's hypothesis (Gale, 1973) that relationshipsbetween E and electrocortical arousal were similarly mediated has not been clearlysubstantiated (Matthews and Amelang, 1993). However, comparable e�ects have been obtainedin evoked potential data (Bartussek et al., 1996; Brocke et al., 1997). The problem with theTMI hypothesis is a calibration problem across studies. The level of stimulation needed toinduce inhibition is never speci®ed a priori and there is little agreement between studies. Brockeet al.'s data on noise and ERPs suggest TMI operates at the moderate intensity of 60 dB. Incontrast, Fowles' study of EDA implies that the much higher intensity of 103 dB (withadditional stress) is necessary before TMI becomes evident in introverts. Fourth, there is littleevidence that N relates consistently to either arousal or arousability, even under stressfulconditions (Fahrenberg, 1987; Naveteur and Freixa-i-Baque, 1987), although there are sporadicpositive ®ndings (Eysenck, 1994) and complex, uninterpretable interactive e�ects. Fifth, therehave been few attempts to test Gray's theory directly and the evidence is mixed. Argumentsthat Imp is particularly predictive of psychophysiological functioning (e.g. O'Gorman andLloyd, 1987) must be balanced against contrary ®ndings (e.g. Swickert, 1996). The attempt byFowles (1980) to link a.n.s. functioning to Gray's theory has been largely unsuccessful,although the De Pascalis et al. (1996) heart rate data provide partial support. By and large,Anx does not relate to increased EDA and Imp does not relate to increased heart rate.Bartussek's ERP studies also fail to provide consistent support for the theory.On balance, the evidence favours the Eysenck theory, at least for extraversion±introversion,

although reservations may be expressed about the level of inconsistency in the data. However,as noted at the outset, reliability of outcome may be reduced by methodological problems anddi�culties in establishing clear conceptual links between speci®c psychophysiological measuresand arousal theory (Eysenck, 1994). Di�ering reactions to these problems are possible. To theextent that consistent results are obtained only in highly selected subjects under narrowly-


constrained testing conditions, the sceptic might reasonably wonder whether psycho-physiological ®ndings tell us much about personality in the wider sense. The contrary view isthat individual di�erences in brain function are more robust than individual di�erencesin psychophysiological response. Eysenck's (1994) view was that: ``Given the enormouscomplexity of the issues involved, the practical di�culties outlined and the instrumentalproblems of working at the boundary of the technically possible, I think we would have beenoptimistic in the extreme had we expected results to have been more positive and in greateragreement with theory'' (p. 198).

4. Mood and subjective states

4.1. A framework for mood studies

Both Eysenck and Gray conceptualise a�ect or mood as a relatively direct output from brainsystems. There are considerable di�culties, however, in establishing mappings between neuralsystems and dimensions of mood due to the complexity of the physiological basis for a�ect(Thayer, 1989, 1996). In reviewing data on personality and mood we shall use the frameworkproposed by Matthews et al. (1990a), developed from Thayer's (1989) work on self-reportarousal. Thayer (1989) has shown that there are two quite distinct bipolar dimensions ofsubjective arousal, which relate to autonomic arousal measures to a similar degree. Energeticarousal contrasts feelings of vigour and tiredness and, according to Thayer (1989, pp. 130±131), relates to Eysenck's reticulo±cortical circuit. This hypothesis is supported by studiesshowing facilitative e�ects of energy on demanding attentional tasks (Matthews, 1992a), in linewith theoretical accounts of cortical arousal and performance (Revelle, 1993). Tense arousal(anxiety vs. calmness) may relate to limbic system arousal, which Eysenck sees as the basis foremotion, especially negative emotion. Matthews et al. (1990a) add a further fundamental mooddimension; hedonic tone or the pleasantness of mood, which contrasts feelings of happinessand contentment with depression and sadness. Links between mood dimensions and brainsystems then support the following predictions. First, introverts should report greater energy:Eysenck (1967, p. 83) stated that ``the concept of fatigue in relation to extraversion±introversion takes the place of the concept of emotion in relation to neuroticism-stability''.Second, N should relate to tension, unpleasantness of mood and mood variability. There is nobasis for relating E to overall pleasantness of mood.Gray (1987a) relates BIS activity (high Anx) to both subjective anxiety and neurotic

depression. Activity of dopaminergic reward systems, such as the BAS, may underpin positivea�ect and subjective energy (e.g. Depue, 1995) and so Imp should tend to relate to positiveemotions. Presumably, noradrenergic arousal heightens the experience of a�ect. Fig. 2 showsthe dynamic interaction between BIS, BAS and the separate arousal system. Matthews et al.(1990a) showed that energy and tension correlated with autonomic arousal but hedonic tonewas independent of autonomic indices. Plausibly, therefore, energy relates to the integratedoutput of the BAS and the arousal system, tension to the BIS and arousal and hedonic tone(HT) to the balance between the mutually inhibitory BIS and BAS systems. In terms of theEysenck dimensions, N should correlate with tension and unpleasant mood (negative HT),


especially when the BIS is activated, whereas E should relate to energy and pleasant mood,

especially when the reward system is activated.

Secondary predictions may be made from (1) Gray's rotation of axes and (2) interaction of

systems in the conceptual nervous system shown in Fig. 2. Because introversion is associated

with Anx, we expect introversion to relate, modestly, to negative emotions, i.e. increased

tension and lower HT. Similarly, the association between Imp and N implies high N subjects

should be somewhat more prone to positive hedonic tone. In addition, Fig. 2 shows that the

BIS and BAS are mutually inhibitory, but both systems tend to excite the arousal system.

Hence, N should relate to suppression of the BAS by the BIS and lower HT. The association

between N and pleasantness of mood may then be somewhat unstable, consistent with the high

mood variability of neurotics (Eysenck and Hepburn, 1989). Presumably, the dominant e�ect

of N varies with motivational context: N should be more positively related to pleasant moods

in situations which do not activate the BIS.

In summary, the two theories make similar predictions concerning a�ective correlates of N,

although the Gray theory also allows for variation of the association between N and positive

mood, depending on the dynamic interaction of the systems. However, the Eysenck and Gray

theories make con¯icting predictions concerning the direction of the relationship between E

and energy (Matthews et al., 1990a). The Gray theory also predicts that E should be associated

with higher HT and perhaps also with lower tension, via the two secondary e�ects described. A

di�culty for the Eysenck theory is that extraverts' stimulation-seeking tendencies and mood-

regulation strategies might compensate for their lower alertness, especially in naturalistic

studies. Extraverts report desiring more activated and more pleasant mood states than do

introverts (Rusting and Larsen, 1995) and extraverts are more likely than introverts to use

exercise to increase feelings of vigour (Thayer et al., 1994).

Fig. 2. Possible inter-relationships between brain systems, mood states and personality traits, within Gray's (1987a)theory.


4.2. Correlational studies

There are two distinct branches of research on personality and mood, which relate topsychometric and experimental traditions respectively. Illustrative results from some of thelarger-scale investigations are shown in Table 1. The ®rst four studies were concerned withestablishing relationships between personality and mood as psychometric constructs: theWatson and Clark (1992) positive and negative a�ect dimensions are similar to energy andtension, respectively. Typically such studies use a diary-based approach or require collegestudents to complete questionnaires for course credit. The studies show consistent correlationsbetween (1) E and energy/positive a�ect and (2) N and tension/negative a�ect, but correlationmagnitudes vary. Meyer and Shack (1989) and Watson and Clark (1992) have argued for aparticularly close correspondence between these pairs of dimensions. Other psychometrically-oriented studies suggest that both E and N relate to overall pleasantness of mood (HT).Williams (1989) found that an elation scale correlated at 0.30 with E and ÿ0.20 with N anddepression correlated at ÿ0.31 with E and 0.45 with N (N=172). The remaining studies listedtook mood measures in the context of experimental studies of performance and show

Table 1Data from illustrative studies of personality and mood

Study N Measures Energy (PA) Tension (NA) Hedonic tone(happiness)

N E N E N E

Costa and McCrae (1980) 575 EPI ÿ11** 16** 35** ÿ01 ÿ ÿPAS/NAS

Emmons and Diener (1986) 72 EPI ÿ02 34** 32** 24** ÿ ÿACL

Meyer and Shack (1989) 231 EPQ ÿ19* 50** 54** ÿ11 ÿ ÿACL

Watson and Clark (1992) 532 NEOÿPI ÿ25** 62** 52** ÿ21** ÿ ÿPANAS

Larsen and Ketelaar (1991)1 70 EPQ ÿ03 10 29** ÿ12 ÿ ÿACL

Adan and GuaÁ rdia (1997)2 578 EPI ÿ20** 18** 26** ÿ08 ÿ27** 24**

UMACLAdan and GuaÁ rdia (1997)3 381 EPI ÿ06 11 18** ÿ01 ÿ24** 23**

UMACL

Matthews et al. (1990a) 158 EPI ÿ25** 13 23** ÿ15 ÿ24** 12UMACL

Matthews et al. (in press) 762 EPQ-R ÿ13* 14* 38** ÿ16** ÿ26** 12**

UMACL

1Neutral mood induction.2Additional data (personal communication, 30/6/97): subjects tested at 9 AM.3Additional

data (personal communication, 30/6/97): subjects tested at 9 PM.Correlation coe�cients multiplied�100, *p<0.05,**p<0.01.PA=positive a�ect, NA=negative a�ect, EPI=Eysenck personality inventory, EPQ(-R)=Eysenckpersonality questionnaire(-revised), PAS=positive a�ect scale, NAS=negative a�ect scale, PANAS=positive and

negative a�ect schedule, UMACL=UWIST mood adjective checklist, ACL=unpublished adjective checklist.


considerably weaker personality±mood associations. The data suggest that N is modestlyrelated to both higher tension and lower energy and to more depressed mood (lower hedonictone). E shows a weak but general tendency to be related to better mood.Several factors may account for variation in results across studies (Dorn and Matthews,

1995). The ®rst is the time-frame of the mood rating, with longer time-frames giving strongercorrelations. Meyer and Shack's (1989) ``state'' measure asked for a mood rating for the pastday, for example, whereas the measure used by Matthews et al. (1990a); Matthews et al., inpress) emphasises feelings `àt the moment''. Second, the in¯uence of personality on moodvaries across situations (BrandstaÈ tter, 1994). Studies of college students in which the ``task'' isno more than the completion of a set of questionnaires may simply pick up characteristicreactions to university life. The happiness of extraverts may in part re¯ect their greaterparticipation in social activities, which appears to be mood-enhancing (Watson et al., 1992).Similarly, N relates to a greater frequency of negative life events (Bolger and Schilling, 1991),although the relationship between N and emotional distress is robust in controlled settings(Matthews and Deary, 1998). Thirdly, the personality measure used may in¯uence correlations.The NEO-PI used by Watson and Clark includes two explicitly a�ect-laden ``facets'' (positiveemotions and activity). In one of their samples, Watson and Clark showed that omitting thesetwo facets in scoring E reduced the E±PA correlation from 0.54 to 0.42, a substantial decreasein the variance explained. Broadly, the correlational data suggest that neurotics are prone tonegative moods and less consistently, extraverts tend to experience positive moods.

4.3. Experimental studies

Experimental evidence on individual di�erences in mood response to manipulated events islimited. Blackburn et al. (1990) used the Velten technique to induce state depression and foundthat N (but not E) predicted the increase in depression. Larsen and Ketelaar (1991) obtained asimilar result when subjects were required to imagine negative scenarios vividly. They alsoshowed mood enhancement due to imagining positive events was stronger in extraverts than inintroverts. Correlations from a neutral mood induction condition are given in Table 1 above.With a negative mood induction, the N±NA correlation increased slightly to 0.34, whereas asigni®cant correlation of 0.32 between E and PA was obtained with the positive induction.Rusting and Larsen (1997) report similar ®ndings.There is a somewhat independent line of research based on interactionist models of anxiety.

There may be multiple anxiety traits related to di�erent contexts, so that N and trait anxietymay only moderate subjective responses to certain types of threat (Endler et al., 1991). Traitanxiety may relate primarily to sensitivity to ego-threat (Eysenck, 1982). Studies conducted inEndler's laboratory (e.g. Busch et al., 1994) suggest that both multiple anxiety traits and theindividual's cognitive appraisal of the nature of the threat must be taken into account inpredicting state anxiety change. Hence, a cognitive explanation for state anxiety change may bemore valid than a neuropsychological one. Neurotic/trait anxious individuals may haveparticularly ready access to negative self-referent information in long-term memory, forexample (Wells and Matthews, 1994) and N relates to cognitive state variables such ascognitive interference, lack of perceived control and low self-esteem in addition to negativemood (Matthews et al., in press).


4.4. Mood: conclusions

N is reliably positively correlated with various aspects of negative mood, consistent withboth Eysenck and Gray theories. N is a considerably more reliable predictor of subjectiveemotion than it is of autonomic arousal, even in the absence of an overt stressor. Hence, it isunclear whether the trait-state relationship should be seen as dependent on neuropsychologicalor cognitive processes (or some interaction between the two). With regard to E, the criticalissue is whether subjective energy is seen as an index of cortical arousal or of BAS activity. Ifwe interpret energy as reticulo±cortical arousal (Thayer, 1989), then its positive correlationwith E is strongly incompatible with the Eysenck theory and unpredicted by Gray. If we seeenergy as a marker for a dopaminergic reward system, its association with E is unpredicted byEysenck, but highly consistent with the Gray theory. The Eysenckian counter-argument is thatextraverts tend to over-compensate for low arousability through stimulation-seeking. Thishypothesis receives some support from the attenuation of the correlation between E and energyin laboratory settings which control for activity preference, but remains somewhat post hoc.There is a marked discrepancy between the tendency of E to correlate negatively withpsychophysiological arousal measures, but positively with subjective energetic arousal.The tendencies for E to relate weakly to reduced tension and N to reduced energy are

also compatible with the secondary predictions made from the Gray theory. The data onmeasures of happiness and hedonic tone suggest that N and E may relate to the balance of theBIS and BAS as Gray predicts. In experimental studies, we expect secondary e�ects of (1)attenuation of negative a�ect response in extraverts, due to their lower Anx and (2)enhancement of positive a�ect response in neurotics, due to their higher Imp. Larsen andKetelaar (1991) and Rusting and Larsen (1997) obtained the ®rst but not the second of thesee�ects. Contrary to Gray's theory, Emmons and Diener (1986) found that Soc was morepredictive than Imp of positive mood. Matthews et al. (1990a) suggested that, in mood studies,it may be preferable to use the Eysenck dimensions, but interpret them in terms of Gray'sdimensions. Overall, mood data ®t the Gray theory quite well, but cognitive mechanismsprovide viable alternative explanations. Matthews (1997) reports that, across two studies,correlations between E and mood covaried with the strength of correlations between E andappraisal and coping scales.

5. Conditioning

Studies of conditioning provide one of the principal sources of evidence on boththeories. Recommended articles in this area include Levey and Martin's (1981) review ofmethodological issues and earlier studies and Corr et al.'s (1995a) thoughtful discussion of theuse of conditioning paradigms to test the two theories. Prediction from Eysenck's theoryis based on the ``drug postulate'', that e�ects of extraversion±introversion on conditioningshould correspond to e�ects of depressant and stimulant drugs respectively. Pharmacologicalstudies indicate that arousal is correlated with rate of conditioning in both associativeand instrumental paradigms, implying that, in general, introverts should condition morerapidly than extraverts (Eysenck, 1994). As ever, testing the prediction requires attention to


task parameters. In particular, in associative learning, TMI may disrupt conditioning ifthe unconditioned stimulus (UCS) is particularly strong, so theory-testing requires the useof a weak UCS. Presumably, relatively weak stimuli are also desirable in instrumentalparadigms.Corr et al. (1995a) contrast the ``Hull±Eysenck'' perspective, that there is a general learning

mechanism, with the ``Mowrer±Gray'' perspective, that there are two learning mechanismsrelated to reward and punishment respectively. The simplest prediction from Gray's (1981)theory is that in anxious individuals, sensitivity to conditioned aversive stimuli leads toenhanced learning with such stimuli. Similarly, Imp relates to learning with conditioned rewardstimuli. However, some complications to these straightforward predictions should bementioned. First, Gray's theory does not deal with generalised sensitivity to reward andpunishment, but with sensitivity to speci®c cues (Zinbarg and Revelle, 1989; Pickering et al.,1995). In conditioning tasks, it is easier to manipulate these cues than in other behaviouralparadigms. Hence, predictions concern conditioning to CSs rather than to USs: Gray (1987a,p. 244) suggests that the Pavlovian association of neutral CSs with unconditioned reinforcersmay be a function of the ®ght/¯ight system. However, Gray (1987b, p. 501) cautions that thedegree of correspondence between individual di�erences in conditioning to CSs and UCSs is anopen empirical question and Gray (1987a, p. 351) makes the explicit prediction that extravertswill condition better with rewarding UCSs.Second, Gray's theory does not specify the role of arousal in conditioning and it is unclear

how the empirical data on drugs and conditioning which provide the basis for Eysenck's drugpostulate are to be explained, unless it is supposed that stimulants act through their e�ects onreward and punishment systems and their e�ects on arousal are incidental.Third, the Gray theory not only proposes two learning factors (reward and punishment

systems), but also two learning processes for associative and instrumental phases of learning(Corr et al., 1995a). Conditioned response to motivationally signi®cant stimuli involves aninitial stage of classical conditioning of a neutral CS to a UCS, followed by a stage ofinstrumental conditioning of the CS to a response. Studies of instrumental learning havetended to con¯ate the two stages, but it may be important to separate them. In particular,anxiolytic drugs fail to in¯uence aversive associative learning (Gray, 1982), so there is norationale for predicting anxiety e�ects on this phase of learning.Fourth, there have been con¯icting statements over whether the theory predicts any form of

learning or not. Until recently, most statements of the theory explicitly related personality tolearning (e.g. Gray, 1981). Corr et al. (1995a, p. 52) state that ``...Gray's theory predicts thattwo separate personality factors should a�ect associative learning under appetitive and aversivestimuli...''. However, Gray (personal communication, 18/11/97) states that the animal studiesof anxiolytic drugs do not support predictions concerning anxiety e�ects on associativeconditioning. Instead, the theory is primarily concerned with the motivating properties ofassociatively conditioned stimuli when the subject has learned a task to an asymptotic level(Pickering et al., 1997). In fact, many human conditioning studies show personality e�ects onlearning following reinforcement cues, indicating a need for further theoretical restructuring, asPickering et al. (1997, pp. 53±53) accept. Furthermore, animal studies of instrumental learninghave shown e�ects of anxiolytic drugs on acquisition but not asymptotic performance(McNaughton, 1985), although these acquisition e�ects relate to response suppression rather


than to conditioning per se (McNaughton, personal communication, 17/4/98). The problem isreally that conditioning is not a basic process, but the outcome of various factors includingsensitivity to motivational signals and pre-existing response tendencies. In this section, we willfocus on the version of the theory (Gray, 1981) actually addressed by most publishedconditioning studies, which predicts individual di�erences in learning. We will also highlightthe key aspects of Pickering et al.'s (1997) comments on these studies.Overall, the strongest test of the Gray (1981) theory would be provided by studies of

instrumental conditioning to weak conditioned stimuli. Under such circumstances both theEysenck and Gray theories predict superior conditioning in passive avoidance and extinctionparadigms in introverts and, in stressful conditions, in individuals high in N. Within Gray'scurrent ``308 rotation'' model, we would expect N to be more strongly related than introversionto BIS-mediated conditioning. However, with BAS-mediated conditioning, in approach andactive avoidance paradigms, Eysenck continues to predict superior conditioning in introvertsand stressed neurotics, but Gray (1981) predicts superior conditioning in extraverts and stableindividuals (especially extraverts). This clear predictive di�erence has inspired much recentwork on conditioning.

5.1. Associative learning

The most popular associative conditioning paradigm has been the learning of theconditioned eyeblink response, in which the UCS is an aversive air-pu�. It seems well-established that with a weak UCS, introverts do indeed condition faster than extraverts, aspredicted by Eysenck (Levey and Martin, 1981). Extensive work on anxiety was conductedwithin a drive theory framework in the 1950s and 1960s, although interest in this topic declinedsharply in subsequent decades. Anxious individuals condition faster than stable individuals instressful circumstances, such as those induced by noxious stimuli or ego-involving instructions(Spence, 1964; Eysenck and Eysenck, 1985). These ®ndings are broadly consistent with theGray theory, to the extent that the stressors concerned activate the BIS. The reliability of theextraversion e�ect is perhaps surprising, given that in Gray's current model E is only weaklyrelated to Anx.However, more detailed scrutiny of individual studies often turns up ®ndings which are

di�cult for either theory to accommodate, such as complex interactive e�ects of E, N andexperimental parameters (Levey and Martin, 1981). E�ects of E on eyelid conditioning areassociated with narrow Imp and venturesomeness rather than with Soc or E per se (Eysenckand Levey, 1972; Frcka and Martin, 1987). Gray (1981) infers that Imp rather than E may bethe key causal factor. However, the result does not seem to support Gray's theory either, inthat Imp should a�ect conditioning only with rewarding stimuli and eyelid conditioningparadigm is usually seen as aversive. Frcka et al. (1983) showed a moderating e�ect of P in aneyelid conditioning study. E was negatively related to CR frequency in low P individuals.Among those high in P, extraverts emitted more CRs. Assuming that P is negatively related tocortical arousal (Eysenck, 1994) the E� P interaction is in the wrong direction. Introverts'superior conditioning should be more reliable in the high P individuals, because of their lowerarousal. Another study providing somewhat ambiguous results is that of Barrett (1971), whoshowed that both low Imp and high Anx were correlated with stronger conditioning. The data


show e�ects of the Gray dimensions, but the problem for the Gray theory is that, with aneutral CS and aversive UCS, it is unclear why BAS activity and Imp should in¯uence classicalconditioning.Corr et al. (1995a) point out that the rather few studies of personality and associative

conditioning which have used appetitive stimuli have yielded mixed results. They cite twostudies which showed no e�ect of E on appetitive conditioning and two studies of sexualconditioning which supported Gray's theory in showing that E was positively related toconditioning with (presumably appetitive) sexual stimuli. Paisey and Mangan (1988) report astudy of electrodermal CRs employing both appetitive (sexual) and aversive stimuli, furthersubdivided into weak and strong stimuli. They concluded that E relates negatively toacquisition of the CR with weak appetitive stimuli and positively to acquisition with strongappetitive stimuli, although results rest on a factor analysis of response indices conducted withan inadequate sample size (31). The role of stimulus strength is similar to that found ineyeblink conditioning (Levey and Martin, 1981) and, in this respect, the ®ndings are consistentwith the Eysenck theory. It is di�cult for Gray's theory to explain why introverts should showstronger conditioning than extraverts with weak sexual stimuli. However, given that extravertsreport more frequent and more diverse sexual encounters than introverts (Eysenck andEysenck, 1985), it is questionable whether sexual responses are ideal for theory-testing:personality is confounded by prior learning. Evidence on the roles of Anx and N on appetitiveclassical conditioning is even weaker. Paisey and Mangan (1988) cite three studies showing apositive, a negative and no relationship between Anx or N and conditioning. Their own studyfound N was negatively correlated with CR formation to appetitive stimuli, irrespective ofstimulus strength, a result which ®ts neither theory, especially if it is assumed that N is amarker for high Imp.Corr et al. (1995a; Experiment 2) used a task which required subjects to learn associations

between coloured lines (neutral CSs) and a subsequent visual display. The subject's responsewas followed by an appetitive, neutral or aversive UCS, indicating loss or gain of money. Ewas negatively correlated with learning in aversive, but not neutral or appetitive, conditions.Corr et al. argue that this ®nding con¯icts with Eysenck's theory, in that personality e�ectswere moderated by type of reinforcement. From the perspective of Gray's theory, it issurprising that it is E rather than N which moderated aversive learning, although there is anargument that associative conditioning is outside the scope of Gray's theory (Pickering et al.,1997).

5.2. Instrumental conditioning

Most studies of instrumental conditioning have used both positive and negativereinforcement. The two most popular paradigms are the verbal operant paradigm anddiscrimination learning of simple psychomotor responses. In verbal operant conditioning, thesubject is reinforced for producing attributes of sentences (e.g. `Ì'' or ``we''), within approachor extinction paradigms. This task has been used mainly to investigate E, though results havebeen somewhat mixed. Mangan's (1982) review of early work suggests that introverts tend tocondition more strongly than extraverts under rewarding conditions, consistent with Eysenck'stheory. However, several studies carried out by B. S. Gupta and his associates consistently


showed that extraverts condition better with a rewarding stimulus, but introverts conditionbetter with a punishing reinforcer (e.g. Gupta and Shukla, 1989). These ®ndings ®t the Graytheory better than the Eysenck theory. Gupta and Gupta (1984) also manipulated arousalthrough administration of D-amphetamine. Results with a punishing reinforcer (electric shock)showed that introverts conditioned more in a placebo condition, but amphetamine improvedconditioning in extraverts and impaired it in introverts. This ®nding is consistent with theEysenck theory, assuming TMI operates in introverts. It does not directly address the Graytheory, given that shock is an aversive UCS rather than a CS. With a positive reinforcer(experimenter says ``good''), extraverts showed greater conditioning than introverts in aplacebo condition. Amphetamine reduced conditioning in extraverts, but had no e�ect onintroverts. Placebo condition data are more consistent with Gray than with Eysenck, but thedrug e�ect is di�cult for either theory to explain. In Gray's theory, both E and amphetamine(as a dopaminergic agonist) should activate the BAS, with the combination of factors leadingto particularly good conditioning. If, as Eysenck claims, extraverts are under-aroused initially,amphetamine should improve their conditioning. Comparison of Imp and Soc in some of thesestudies has shown that both dimensions give similar results to analyses of E (Gupta, 1990).Two studies using discrimination learning are of particular relevance to theory (Zinbarg and

Revelle, 1989; Corr et al., 1995a). Zinbarg and Revelle point out that moderation of learningby cue type provides a more direct test of Gray's theory than moderation by type ofreinforcement. They required subjects to learn to respond to ``go cues'' or reward signals,indicating reward or active avoidance of punishment and to ``no-go cues'' or punishmentsignals, indicating omission of reward or passive avoidance of punishment. Reinforcement type(reward vs. punishment) was also manipulated: subjects could win or lose points. Gray's theorypredicts interactive e�ects of Imp and Anx with cue type. The study was designed not to beover-arousing, so Eysenck's prediction of generally better conditioning in introverts might betested. Across four studies, the most consistent learning e�ect was the interaction between cuetype, Imp and Anx. With go cues, high Anx facilitated learning among low impulsives, butwith no-go cues, Anx facilitated learning among high impulsives. This pattern of interaction isinconsistent with Gray's theory, most simply because Anx should not in¯uence conditioning togo cues and Imp should not a�ect performance with no-go cues. Zinbarg and Revelle modifyGray's model to accommodate their ®ndings by incorporating constructs of expectancy andprocessing resources, although they admit the modi®cations are speculative. E and N showedsomewhat similar results to Imp and Anx respectively, but ®ndings with these dimensions wereless consistent, implying that the Gray dimensions may be more valid as predictors ofconditioning. Results provided little support for the Eysenck theory, with introverts showingno general superiority in conditioning.Corr et al. (1995a; Experiment 3) ran a somewhat similar discrimination learning study in

which subjects could win or lose small sums of money on each trial of a psychomotor task, bymodulating response speed following presentation of passive avoidance and approach cues.Speed of response to approach cues was related to lower trait anxiety and frequency ofpunishments in the passive avoidance condition was higher in impulsive subjects. These resultsdo not clearly ®t either theory. From Gray's perspective it is hard to see why Anx shouldrelate to approach behaviour, which is mediated by the BAS and Imp should relate to BIS-mediated passive avoidance. The authors suggest the ®nding should be attributed to


inconsistency of response suppression during punishment. The interaction between Imp, Anxand type of reinforcement obtained by Zinbarg and Revelle (1989) was also tested and foundto be non-signi®cant.Pickering et al. (1997) describe the unpredicted e�ects of anxiety and impulsivity in the Corr

et al. (1995a) study as ``complementary-trait'' e�ects. Similar ®ndings suggesting that rewardsignals increase behavioural activation in low anxious subjects have also been found in maze-crossing and card sorting tasks and in a startle re¯ex study whose results failed to replicatesubsequently (see Kumari et al., 1996; Pickering et al., 1997). It is suggested that, in moreanxious individuals, individual di�erences in BIS activity elicited by the experimental settingtend to mask e�ects of positive reinforcement, because of the inhibitory e�ects of the BIS onthe BAS. This hypothesis does not explain the absence of e�ects of anxiety on passiveavoidance in the Corr et al. (1995a) study. Pickering (1997) suggests that neural networkmodelling of the interaction between BIS and BAS systems may resolve anomalous ®ndings. Asimple mutual inhibition model failed to show the complementary-trait e�ect. A further modelincluded an arousal unit in the network and made the radical proposal that the BIS and BASa�ect response only via their facilitative e�ects on arousal. If it is assumed that (1) the BIS hasa stronger e�ect on arousal than the BAS and (2) arousal is related to response speed by theYerkes±Dodson Law, then the complementary-trait e�ect for anxiety can be modelled. Giventhe further modi®cation of the revised Gray theory here, it is perhaps safest to view this resultas a demonstration of the potential of neural network modelling for explaining unexpectede�ects, rather than as a strong explanation for the empirical ®ndings.Various studies using other learning tasks show superior learning of extraverts under reward.

Pickering et al. (1995) list these studies and ®nd general support for greater sensitivity toreward in extraverts and greater sensitivity to punishment in introverts. However, they pointout a variety of di�culties in deriving clear predictions for learning tasks and detailed taskanalysis may be necessary. There is evidence too for arousal e�ects on learning. Corr et al.(1995b) report that ca�eine enhances procedural learning in extraverts, but impairs learning inintroverts, consistent with other performance research reviewed in the next section. This e�ectwas associated with Soc, but not Imp.

5.3. Conditioning: conclusions

First, ®ndings are only consistent within more ``traditional'' conditioning paradigms, notablythe eyeblink and verbal conditioning paradigms. It is disturbing that the well-designed andthorough discrimination learning studies of Zinbarg and Revelle (1989) and of Corr et al.(1995a) give entirely di�erent results. In discrimination learning, subjects are explicitlyinstructed to learn, which may make the task more dependent on high-level cognitivemechanisms. Second, eyelid conditioning and verbal instrumental conditioning underpunishment are related to personality broadly as the Eysenck theory would predict, althoughthere are some anomalies in speci®c studies (Levey and Martin, 1981; Frcka et al., 1983).Enhanced learning of introverts under these conditions may be explained by the Gray (1981)theory to the extent that introversion is a proxy for high Anx. Third, results with verbaloperant conditioning under reward are more consistent with Gray than with Eysenck. Fourth,some studies suggest that Anx and Imp are more reliable predictors of conditioning than E and


N, but others fail to support this conclusion. Fifth, the most recent account of Gray's theory(Pickering et al., 1997) muddies the waters by eschewing prediction of individual di�erences inlearning and by emphasising interesting but unpredicted empirical ®ndings such as``complementary-trait'' e�ects. It is di�cult to evaluate post hoc interpretations for such e�ectswithout further empirical work.On balance, ®ndings appear to favour the Gray (1981) theory, in that the Gupta studies

con®rm a primary prediction of the theory, enhanced instrumental conditioning to rewardsignals. However, di�erent reinforcers may vary in their e�ects on arousal and personality maymoderate the arousal response to motivationally signi®cant stimuli, so that manipulationsintended to test the Gray theory are confounded with arousal. The argument is sound, buttends to be used in a rather inconsistent way. For example, Eysenck and Eysenck (1985, p.245) suggest that appetitive conditioning may be more arousing than aversive conditioning (atleast with sexual stimuli). Corr et al. (1995a, p. 52), however, supposed that reward is lessarousing than punishment. They showed that the aversive CS induced more subjective tensearousal on the UWIST Mood Adjective Checklist (Matthews et al., 1990a) than did theappetitive CS. The ®nding that introversion related only to conditioning with an aversive CS isthus problematic for the Eysenck theory, because introversion predicted conditioning only inthe most arousing of the three reinforcement conditions used. If (non-sexual) rewards onlyinduce moderate arousal, the superiority of extraverts on verbal operant conditioning underreward becomes even more di�cult to explain from the Eysenck theory. It is unfortunate thatso few studies have tested for possible confounds.Some of the data may be explained by combining aspects of the Eysenck and Gray theories,

supposing that (1) arousability of extraverts and introverts is moderated by the incentivevalue of the stimulus and (2) arousal is the principal in¯uence on conditioning. If, as thesubjective arousal data suggest, extraverts are more aroused by reward signals (or possiblyreward stimuli in general), then extraverts may show superior conditioning in rewardingconditions simply because they are more aroused than introverts. Impairment of conditioningin extraverts given amphetamine (Gupta and Gupta, 1984) might then re¯ect TMI, although itis curious that the apparent interaction between E and stimulus strength in conditioning withsexual stimuli suggests greater TMI in introverts (Paisey and Mangan, 1988). Similarly,introverts may be more aroused than extraverts under punishment conditions and during eyelidconditioning.

6. Attention and performance

Studies of human performance have long been of interest to researchers testing Eysenck'sarousal theory. In recent years, there has been an upsurge of interest in the implications ofGray's theory for performance. Eysenck links personality to performance on the basis of threequite separate hypotheses. First, introverts and stressed neurotics tend to be chronically high incortical arousal during performance. Second, cortical arousal is related to performance by theinverted-U curves speci®ed by the Yerkes±Dodson Law. Third, subjects are motivated to seekan intermediate level of arousal, which may a�ect their strategy for response. Hence, e�ects ofE on performance should be moderated by arousal. In particular, the normally under-aroused


extravert should perform more e�ciently under stimulating conditions, whereas de-arousingenvironments favour the introvert. Gray's theory makes fewer predictions outside theconditioning paradigms already discussed. Much of the work to be reviewed is concerned withthe implications of ``sensitivity'' to reward and punishment cues for performance on taskswhich do not correspond closely to those used in animal research. It is often unclear whethersensitivity to motivational signals should be expressed as performance enhancement or asdistractibility.The literature on personality and performance is far too extensive to review in full (see

Eysenck and Eysenck, 1985; Matthews, 1992b, 1997; Matthews and Dorn, 1995). In thissection, we consider key studies in four areas of particular theoretical relevance: the interactionof extraversion and arousal, attentional systems, memory and responsiveness. These studieshave focused mainly on E. In the ®nal sub-section, we brie¯y review the theoreticalimplications of studies of Anx and N conducted outside the Eysenck and Gray frameworks.

6.1. Extraversion: moderating e�ects of arousal and task factors

The simplest prediction from Eysenck's theory, that extraverts should tend to out-performintroverts in arousing conditions, is quite well-supported (Eysenck and Eysenck, 1985). Studiesof perceptual thresholds provide striking evidence. Introverts show higher auditory and visualthresholds than extraverts in baseline conditions, but additional stimulation in a di�erentmodality leads to threshold increasing in introverts, but decreasing in extraverts (Shigehisa andSymons, 1973; Shigehisa et al., 1973). Extraverts' superiority in performance on attentionallydemanding tasks (other than vigilance) noted by M. W. Eysenck (1981) is also consistent withtheory. According to the Yerkes±Dodson Law, more di�cult tasks have a lower optimal levelof arousal for performance and so extraverts should be advantaged. At one level, this line ofresearch supports the Eysenck theory, in that Gray's theory provides no straightforwardexplanation for these performance e�ects.However, the assumptions made in linking data on E�arousal interactions to arousal have

been increasingly challenged. Typical factorial designs are notoriously weak as a means fortesting arousal theory (Hockey, 1984) and, as discussed previously, the evidence for highertonic arousal in introverts in performance assessment settings is unconvincing. At an empiricallevel, three areas of di�culty for arousal theory have emerged: the role of time of day, the roleof task factors and the lack of evidence for arousal-mediation. Interaction between E andarousal appears to vary with time of day (Revelle et al., 1980). In the evening, extravertsperform better under low arousal, implying they are over-aroused. Revelle et al. (1980) suggestthat e�ects of E are driven by Imp, which relates to a phase di�erence in circadian arousalrhythm, such that high impulsives are indeed more aroused than low impulsives in the evening.There are di�culties for this theory too, such as the weakness of the association between E andmorningness-eveningness (Matthews, 1988; Tankova et al., 1994). The reliability of thesuperiority of Imp over Soc measures as predictors of performance has also been questioned(Amelang and Ullwer, 1991). Nevertheless, the Eysenck theory o�ers no explanation for themoderating e�ect of time of day.Another type of moderating factor is the nature of the task. The Yerkes±Dodson Law

concerns generalised cortical e�ciency, such that task di�culty but not the precise information-


processing demands of the task in¯uence the arousal-performance relationship. In fact, taskdemands appear to be of critical importance. The characteristic E� arousal interaction islargely restricted to simple tasks requiring encoding of easily-perceived stimuli or to morecomplex tasks with a routine encoding component (Matthews and Dorn, 1995; Matthews,1997). Attentionally demanding tasks do not show the e�ect, although the Yerkes±DodsonLaw predicts that the over-arousal of introverts should be particularly deleterious on di�culttasks. Matthews and Harley (1993) relate the interaction speci®cally to spreading activationprocesses related to low-level encoding.A ®nal di�culty is evidence that e�ects of E are not in fact mediated by individual

di�erences in cortical arousal. Matthews (1992b) reviews studies in which measures of allrelevant constructs were taken Ð arousal, E and performance Ð and concludes that theyprovide little support for the arousal-mediation hypothesis. In studies of self-report arousal,interactions between E and arousal show e�ects of E with arousal statistically controlled(Matthews, 1985). In other words, arousal moderates rather than mediates the E e�ect. Similar®ndings have been obtained when arousal is measured through EEG alpha power (Matthewsand Amelang, 1993).

6.2. Studies of attention

Attentional paradigms are frequently used in personality research, but many studies neglectthe distinctions between di�erent systems developed by cognitive neuropsychologists (e.g.Posner and Rothbart, 1991). The Eysenck theory predicts a general enhancement of attentionin more aroused individuals (i.e. introverts), but more speci®c predictions may be derived forthe Gray theory.

6.2.1. Spatial attentionOrienting attention in space is a functionally and anatomically distinct aspect of attention.

Typical experimental studies use a cue to draw the subject's attention to a peripheral location,followed by a target which may or may not be at the cued location. Posner and Rothbart(1991) describe a ``posterior'' orienting system which is innervated by the DNAB, such thatrelease of noradrenaline facilitates speed of orienting and slower disengagement from theattended stimuli (Posner and Raichle, 1994). Gray's theory implies that impulsives givenreward signals and anxious individuals exposed to punishment cues should show greaterDNAB arousal and enhanced spatial orienting. Consistent with prediction, Avila (1995)and Derryberry and Reed (1997) showed that N and Anx are associated with slowerdisengagement of attention following orienting to a cued location in space. In Derryberry andReed's (1997) study the e�ect was contingent upon most trials being negative. However, Anxfails to predict speed of moving to a cued location, even when cues are negative (Derryberryand Reed, 1997).Derryberry and Reed (1994) used two target locations, one associated with reward and one

with punishment (gaining or losing points). They argued that Gray's theory predicts thatpersonality should interact with cue type (i.e. cue at positive location vs. negative location).They also manipulated post-trial feedback. Results were complex, but the most consistent e�ectacross the various studies was an interaction between E, cue type and feedback on the previous


trial. Personality e�ects were strongest following negative feedback. Under these conditions,extraverts were relatively slow following a positive cue, but introverts tended to show greatercosts of the negative cue. These e�ects tended to be stronger in high N individuals. Derryberryand Reed (1994) suggest that disengagement is slowed in neurotic introverts (high Anx) givennegative feedback and in neurotic extraverts (high Imp) given positive feedback.Results from spatial attention studies provide some support for Gray, but there are problems

also. First, Gray's theory does not provide a rationale for the moderating role of feedback,which is an actual outcome rather than a signal of reinforcement (Derryberry and Reed, 1994).Second, Derryberry and Reed (1994) report brie¯y that analysis in terms of E and N provideda more coherent pattern of results than analysis using Imp and Anx factors. Third, the e�ectseems speci®c to disengagement; Gray's theory provides no rationale for distinctions betweenthe di�erent components of spatial orienting. It is di�cult to accommodate ®ndings on spatialattention within the Eysenck theory in any simple way.

6.2.2. Target detectionIt is often supposed that the analysis of a stimulus required for object identi®cation or

discrimination (``what is it?'') is distinct from analysis of its spatial location (``where is it?'').Posner and Raichle (1994) describe an anterior system for detection of events located in areasof midprefrontal cortex, in¯uenced by dopaminergic a�erents. It is active during visual targetdetection and conscious e�ortful control of attention.Tasks in which frequent targets are delivered through multiple channels are expected to

activate the anterior system. In fact, extraverts seem to enjoy a general performance advantageon multiple-channel tasks, especially with verbal or symbolic stimuli (M. W. Eysenck, 1981;Matthews, 1997). This e�ect does not appear to be arousal-mediated: Eysenck and Eysenck(1979) demonstrated superior dual-task performance in extraverts irrespective of manipulatedarousal (noise). Tentatively, this e�ect might be attributed to dopaminergic in¯uences onPosner's anterior system, consistent with Gray's hypothesis that Imp (and hence E) correlateswith dopaminergic activity. There appears to be anatomic overlap or near-overlap between theparalimbic structures, such as parts of the cingulate cortex, which contribute both to the eventdetection system and to the BIS function of inhibition of motor plans. Furthermore, e�ects ofsubjective energy on demanding attentional tasks are often considerably stronger than those ofE and energy may re¯ect activity of dopaminergic circuits (Matthews and Davies, 1998).Anx and N frequently impair demanding, non-spatial attentional tasks (e.g. M. W. Eysenck,

1992). However, psychobiological explanations may be misplaced. Performance de®cit isassociated with worry rather than with emotion and bodily symptoms and anxiety interactswith a variety of information-processing parameters in ways unpredictable from arousaltheory, leading most researchers to adopt cognitive theories of anxiety e�ects (M. W. Eysenck,1992). Derryberry and Reed (1997) have shown enhanced attentional focusing to targets innegative incentive conditions, although, in contrast to spatial attention, there were nomoderating e�ects of the incentive value of the individual trial or of feedback. The e�ect wasfound only for right visual ®eld targets, supporting an explanation in terms of ``tonicactivation'' of dopaminergic projections from the ventral tegmental area to left hemisphereobject processing pathways. This account of the anxiety e�ect is at odds with Gray'sconception of anxiety.


6.2.3. VigilanceThe traditional vigilance task requires detection of infrequent targets over prolonged periods

of work. Arousal tends to fall during performance and so, according to the Eysenck theory,extraverts should be particularly prone to under-arousal and performance decrement. Posnerand Rothbart (1991) relate the DNAB to a vigilance function, the clearing of consciousness tomaintain readiness for detecting incoming stimuli. Impulsives and high anxious individuals aremore prone to arousal of the DNAB (see Fig. 1) and so these two groups should tend to showsuperior vigilance. Vigilance tasks tend to be somewhat unpleasant, so we might expect Anxe�ects to predominate. In this case, the evidence favours Eysenck over Gray, at leastsuper®cially. Introverts show an advantage in detection rate and perceptual sensitivity onvisual, but not auditory, vigilance tasks (Koelega, 1992) and extraverts' performance de®cit issometimes eliminated by arousal manipulations (Davies and Parasuraman, 1982). Giambra etal. (1989) found that such e�ects relate to Soc rather than Imp. Unfortunately, results fromstudies which have taken an independent measure of arousal suggest that e�ects of E are notdirectly mediated by individual di�erences in cortical arousal (Matthews, 1992b). High eventrate tasks used in recent research are sensitive to arousal, but rarely sensitive to E (Matthewsand Davies, 1998). The prediction from Gray's theory that Anx should enhance vigilance is notsupported. Davies and Parasuraman's (1982) review concludes that e�ects of N are ``minimal''and state anxiety may actually impair vigilance (Geen, 1985). An alternative means forinvestigating Posner's vigilance system is through assessing the phasic alertness generated by awarning signal in RT studies. Derryberry (1987) predicted that, according to Gray's theory,introverts should show enhanced alerting to a negative warning signal, but failed to ®nd anye�ect of E on alerting processes across two studies.

6.3. Studies of memory

E�ects of E on memory frequently depend on retention interval. Broadly, extraverts showsuperior recall up to intervals of about 5 min, after which introverts show an increasing recalladvantage as retention level increases (e.g. Howarth and Eysenck, 1968). Eysenck and Eysenck(1985) explain extraversion±introversion e�ects on memory on the basis of action-decrementtheory, which claims that high arousal inhibits retrieval in the short-term so as to enhanceconsolidation and long-term retention. As in the case of E� arousal interactions, we have aphenomenon for which the Eysenck but not the Gray theory has a ready explanation.However, as M. W. Eysenck (1981; 1982) discusses, action decrement theory is unsatisfactory,because there are a variety of instances in which arousal e�ects on memory are not aspredicted. As with E� arousal interactions, there are also concerns over the role of time of day(Revelle and Loftus, 1992), task factors (M. W. Eysenck, 1981) and the mediating role ofarousal (Matthews, 1992b). The exact neural basis for personality e�ects on memory is alsounclear.Other personality factors and the emotional content of stimuli may also in¯uence memory.

Recent evidence suggests that e�ects of E on incidental free recall may be moderated bystimulus valence (Bartussek, 1996), but there is a general lack of both theory and data in thisarea. N and Anx tend to be associated with impairment of memory at both short and longretention intervals, a result which is inconsistent with action-decrement theory (M. W. Eysenck,


1981). Contemporary explanations for such e�ects focus on cognitive mechanisms such asreduced elaboration of encoding and allocation of working memory to processing worries(M. W. Eysenck, 1992; Wells and Matthews, 1994).

6.4. Responsiveness

Both theories provide a rationale for personality in¯uencing response rate. According toEysenck (1967), extraverts should be prone to frequent response as a strategy to raise theirarousal towards the hedonic optimal, especially in low arousal environments, whereasintroverts should be reluctant to respond. Stressed neurotics should also avoid response inorder to lower arousal. According to Gray, vigour of response is one of the principal outputsof the arousal system, whereas probability of response relates to the BAS and BIS systems.Because both the BAS and BIS activate the arousal system, both Imp and Anx might beexpected to raise responsiveness, especially under the appropriate reinforcement conditions,with the proviso that Anx e�ects may be countered by behavioural inhibition.Extraverts tend to show faster response on short tasks, riskier speed-accuracy tradeo� and

lower response criterion (Eysenck, 1967). A series of RT studies conducted by Brebner andCooper (1985) suggested that extraverts appear to be ``geared to respond'' and introverts seem``geared to inspect''. However, e�ects of E on responsiveness do not seem to be very reliableacross di�erent tasks and are not apparent in many well-designed studies (Amelang andUllwer, 1991; Matthews, 1992b). Koelega's (1992) meta-analysis of vigilance showed nosigni®cant e�ects of E on false positives and RT, but, in extreme-group studies, responsecriterion (indexed by beta) was lower in extraverts. When e�ects are observed, there issurprisingly little evidence on their relationship to individual di�erences in arousal. Matthewset al. (1990b) found that, in the morning, energetic arousal was positively related toresponsiveness (i.e. low beta) in extraverts but not introverts, with the e�ect tending to reverseor disappear in the evening. As with the performance data previously discussed, theE�arousal interaction suggests that arousal moderates rather than mediates extraversione�ects. Arousal theory also ignores evidence that speed-accuracy tradeo� may be in¯uenced bya variety of independent processing stages. Dickman and Meyer (1988) used a stage analysis ofa visual comparison task to show that Imp did not have a general e�ect on speed-accuracytradeo�, but in¯uenced a feature comparison stage taking place relatively early in processing.RT studies sometimes distinguish ``decision time'', which aggregates stages up to and includingresponse selection and ``movement time'', which refers to response execution. Stelmack et al.(1993a,b) found that E was reliably negatively related to movement time on various speededresponse tasks, but not to decision time. Doucet and Stelmack (1997) replicated this ®nding.Newman et al. (e.g. Wallace et al., 1991) tested a modi®ed version of the Gray theory in

studies of responsiveness. They identi®ed the arousal system with N and the balance of the BISand BAS with E, so that N tends to amplify the reactivity of extraverts to reward stimuli andthe reactivity of introverts to threat stimuli. The dominant response set also moderatespersonality e�ects. In a study of pattern matching, Nichols and Newman (1986) found anoverall tendency for extraverts to respond relatively rapidly in reward-only conditions, but forintroverts to respond more quickly in punishment-only conditions. This e�ect might be seen asmediated by individual di�erences in arousal dependent on the relative sensitivities of BAS and


BIS. Hernaiz-Sanders (1991) failed to replicate the e�ect. She identi®ed two di�culties with theNichols and Newman paradigm. First, it is di�cult to distinguish e�ects of unconditionedreinforcement from reinforcement signals. Second, the consequences for responsiveness of BISactivation are hard to predict, because its outputs include behavioural inhibition as well asarousal. Nichols and Newman also found that extraverts tended to speed up followingpunishment, but introverts tended to be faster after reward, when both reward andpunishments are delivered across trials. As discussed above, Derryberry and Reed (1994)obtained similar e�ects in a spatial orienting paradigm and argued that incentive processes maya�ect ability to disengage from stimuli.N is not related to low responsiveness, contrary to the Eysenck theory. Stelmack et al.

(1993a,b) found that N related to faster decision time, but was unrelated to movement time.State anxiety may lead to a more impulsive strategy under stressful conditions (Leon andRevelle, 1985). Wallace and Newman (1990) demonstrated `ànxious impulsivity'' using a circletracing task requiring continuous response. Response was faster in neurotic introverts (highAnx subjects) than in stable extraverts (low Anx subjects) when goals are uncertain, acondition which may activate the arousal rather than the inhibition response of the BIS. Witha de®nite behavioural goal, neurotic extraverts (high Imp subjects) traced faster than stableintroverts (low Imp subjects). Behavioural goals may activate the BAS. Given Newman'sassumptions about the task, these data seem quite consistent with Gray's theory, although littleattempt was made to consider alternative explanations. The main di�culty is that responsecriterion and response latency are not direct indices of some basic reactivity, but depend onstrategic decisions based on a statistical model of the task performed (Matthews, 1996). Withinsignal detection theory, response vigour (RT) and response probability (beta) are closelyrelated (Davies and Parasuraman, 1982). It is hard to interpret responsiveness data without aninformation-processing analysis of the task.

6.5. Attention and performance: conclusions

In our view, both theories are having di�culties in accommodating the diverse empirical®ndings emerging from di�erent paradigms. The Eysenck theory succeeds fairly well inexplaining some of the general trends in the extraversion data, such as arousal-relatedimpairments in introverts and the responsiveness of extraverts. On some tasks, such as sensorythreshold tasks and paired associate learning, the correspondence between data and predictionis impressive (Eysenck, 1997). However, studies of attention and memory tasks exposes thelimitations of arousal theory, such as the failure of any study of performance to demonstratedirectly that e�ects of E are in fact arousal-mediated (Matthews, 1992b). There are also largeareas of performance research in which the theory either makes no predictions or even basicpredictions from the theory are not supported. Arousal theory, in Eysenck's formulation, doesnot explain personality e�ects on performance in the afternoon and evening (Revelle et al.,1980), on spatial orienting (Derryberry and Reed, 1997), on demanding target detection tasks(Matthews et al., 1990b) and on a variety of memory tasks (M. W. Eysenck, 1992). E�ects ofN in moderating e�ects of E on performance are often broadly compatible with arousal theory(see Eysenck and Eysenck, 1985), but, for the most part, contemporary work on Anx and Nmakes little reference to arousal and focuses on cognitive and attentional mechanisms.


Tests of Gray's theory have been somewhat stymied by di�culties in deriving testablepredictions on typical performance tasks. The theory receives some support from motivationalmoderation of e�ects of E. However, the most reliable e�ect of this kind, associated withindividual di�erences in reactions to feedback on a previous trial, relates to motivationaloutcomes rather than cues (Derryberry and Reed, 1994) and so relevance of the theory isuncertain. Some other ®ndings are broadly compatible with the theory, such as `ànxiousimpulsivity'' (Wallace and Newman, 1990). Speculatively, extraverts' superiority in multi-channel performance and greater responsiveness might be attributed to the BAS anddopaminergic activity, although apparently impulsive response style may actually relate toindividual di�erences in stimulus analysis (Dickman and Meyer, 1988). Gray (e.g. Gray, 1982)deserves credit for predicting the bias in selective attention characteristic of anxiety (Wells andMatthews, 1994) prior to its demonstration in experimental studies. However, there areconsiderable di�culties in mapping BIS functions onto human selective attention functions(Wells and Matthews, 1994, pp. 327±332) and, embarrassingly for Gray's theory, cognitivetherapy abolishes the bias in selective attention in anxiety patients, but anxiolytic drugs do not(Golombok et al., 1991). In addition, attentional bias is found in panic disorder (McNally etal., 1990), a condition Gray (1987a) attributes to the ®ght/¯ight system rather than the BIS.At best, neither theory seems likely to account for more than a subset of observed

personality e�ects on performance. The areas of research to which the theories fail tocontribute are su�ciently extensive that predictive failures cannot simply be attributed toinappropriate choice of task parameters. Furthermore, there are acceptable cognitivemechanisms for explaining many of the ®ndings problematic for the psychobiological theories(e.g. M. W. Eysenck, 1992; Matthews, 1997). If either theory is to have a future role inperformance research, it must either specify improved methods for identifying behaviouralindices controlled by the brain systems of interest or tackle the di�cult problem of howoutputs from those systems may interact with information-processing mechanisms.

7. Conclusions

The evidence reviewed raises several interlocking issues. First, the theories disagree on thekey traits and moderating factors. Eysenck emphasises E, N and level of stimulation, whereasGray focuses on Imp, Anx and motivational signals. We will consider whether the data suggestone set of variables is more predictive of response than the other set. Second, evidence relatingto direct tests of theory will be assessed. Finally, the wider implications for thepsychobiological approach to personality theory will be discussed.

7.1. Indirect tests of theory: choice of constructs and moderating factors

Consistent with Gray's (1981) position, there are some areas of research where Imp doesseem to be somewhat more predictive than either Soc or E, including the spontaneous EEG(Stenberg, 1992) and interactive e�ects of personality and ca�eine on verbal ability (Revelle etal., 1980). However, these Imp-driven e�ects seem very speci®c: for example, the EEG resultfails to generalise to brainstem EPs (Swickert, 1996). Often, di�erences between Imp and Soc


are either absent or inconsistent across studies (Gupta, 1990; Amelang and Ullwer, 1991; Corret al., 1995a). Hence, although Imp may be more predictive in certain paradigms, overall thereis little compelling reason to favour the Gray axes over the Eysenck ones. More generalpsychometric evidence favours the Eysenck and Eysenck (1985) structural model, of course.The other choice between constructs concerns the factors which moderate relationships

between traits and response. Probably the single most secure conclusion from biologically-based trait research is that level of stimulation or arousal moderates associations between Eand response (Eysenck and Eysenck, 1985). The question raised by Gray's theory is whetherthere are moderating e�ects of motivational signals over and above these arousal e�ects.Clearly, motivational manipulations moderate personality e�ects, but in some cases thesemanipulations may operate through their stimulating e�ects (Eysenck and Eysenck, 1985;Bartussek et al., 1996). Other paradigms, notably mood induction and instrumentalconditioning, show di�erential e�ects of reward and punishment manipulations broadlycompatible with Gray's theory. However, it is unclear that reward and punishment cues (CSs)have a stronger moderating e�ect than actual rewards and punishments (USs), as the theoryimplies (cf. Derryberry and Reed, 1994) and it may be di�cult to discriminate US from CS inhuman subjects. Motivational e�ects may be confounded with arousal, but, as we suggested inthe section on conditioning, this argument seems to be used in a rather post hoc way.Personality may moderate the arousing e�ects of motivational stimuli, a possibility which hasreceived insu�cient attention. The data suggest that motivational and arousal-related factorsmay both be important as moderator variables, but no ®rm conclusions may be drawn.

7.2. Direct tests of theory

Predictions from the Eysenck theory of extraversion are supported in several research areas.Overall, the psychophysiological data are suggestive of lower arousability of extraverts in low-stimulation environments and greater arousability of extraverts in high-stimulationenvironments, especially in ERP and EDA paradigms. E is a weak predictor of tonic arousalmeasures. These e�ects are consistent with the theory if we allow that TMI develops inintroverts under high stimulation. In behavioural paradigms, similar e�ects have beendemonstrated across multiple studies in eyeblink conditioning (Levey and Martin, 1981) and insensory threshold studies (Shigehisa and Symons, 1973). Together, these results provide a solidcore of empirical support for the Eysenck theory of extraversion.In other research areas, the Eysenck theory has been less successful. Sometimes E simply

fails to a�ect response measures at all, though such results might re¯ect insu�cient statisticalpower or other methodological problems. More seriously, there are many studies showinginteractions between E and other factors which cannot readily be explained by the theory,although tentative, post hoc explanations are advanced in some cases. Such ®ndings suggestincompleteness of the theory. Such problems are especially apparent in the performancedomain, within which e�ects of E do not seem directly contingent upon individual di�erencesin arousal (Matthews, 1992b). Performance correlates of E are best conceptualised as a``cognitive patterning'', associated with a variety of independent information-processingmechanisms (Matthews, 1997).


The Eysenck theory has been less successful in providing an account of N. It predicts thecharacteristic negative a�ect of neurotics and their greater sensitivity to negative moodinductions. It also predicts superior associative conditioning in anxious individuals. However,there is a striking mismatch between the robust subjective data on N and negative mood andthe inconsistency of associations between N and autonomic arousal, even when stress isexplicitly manipulated. As in the case of E, performance studies provide super®cial support forarousal theory, in that neurotics (or high Anx subjects) are especially disadvantaged on moredi�cult tasks). Again, the arousal theory explanation breaks down when scrutinised in detail,with most contemporary theories of anxiety making worry rather than arousal the centralexplanatory construct (M. W. Eysenck, 1992; Wells and Matthews, 1994).Overall, the Eysenck theory successfully explains an important subset of the evidence, but

not all of it. The operation of TMI is critical for ®tting outcomes to prediction, but we havenoted the problems in calibrating the levels of stimulation necessary for TMI across studies,which prevent ante hoc prediction of TMI. The same problem emerges in comparing resultsacross di�erent areas of research. For example, psychophysiologists assume that performanceof demanding tasks (Fowles et al., 1977) or moderate dosages of ca�eine induces TMI inintroverts (Smith, 1983). But in performance research, similar dosages are seen as arousing inboth extraverts and introverts (Revelle et al., 1980). Basic assumptions about arousal levelsvary radically across studies and so it is di�cult to decide on the paradigms where arousaltheory might or might not be useful.There are two questions for the Gray theory, given the partial success of the Eysenck theory.

First, can it provide an alternative explanation for the arousal-dependent e�ects of E which theEysenck theory accommodates? Second, are ®ndings inconsistent with the Eysenck theorypredicted by Gray's theory? Gray's theory might explain the arousal ®ndings as follows.Introversion is correlated with sensitivity of the BIS, one of whose outputs is arousal. Hence,given that most laboratory environments are likely to provide at least a low level ofpunishment or novelty cues, we expect that BIS activation will feed into higher arousal inintroverts. This hypothesis has two di�culties. Gray's arousal system is supported by theDNAB, but there is little evidence that this system shows any equivalent of TMI, i.e. inhibitionunder high levels of stimulation. Thus, the hypothesis fails to explain the role of stimulationlevel in moderating correlations between E and arousal. Second, the 308 rotation model statesthat the BIS is better indexed by N than by E. Hence, if introverts are typically somewhatmore aroused, due to activation of the BIS, then neurotic individuals should be markedly morearoused than stable individuals, which is simply not the case. The classical extraversion/arousaldata ®t Eysenck's theory better than Gray's, although it might be argued that these e�ects areoutside the scope of the Gray theory.The second issue for Gray's theory is its capacity to explain additional ®ndings. Gray

himself emphasises studies of conditioning to motivational stimuli as a source of evidence, butthe evidence is mixed. The single strongest piece of evidence in favour of Gray's theory derivesfrom the Gupta studies showing that extraverts condition better than introverts to verbalreward. The post hoc explanation o�ered by arousal theory, that verbal reward induces TMI inintroverts seems a little contrived in this instance. However, the e�ect is very paradigm-speci®c:personality e�ects on discrimination learning of positive cues do not provide strong support forGray (e.g. Zinbarg and Revelle, 1989). Other ®ndings which may support Gray's theory are the


associations between E and subjective energy and ``behavioural activation'' performancemeasures such as faster movement time, although in these cases the link between the criterionand the BAS is somewhat speculative.In other studies of motivational moderators, support for the Gray theory is patchy, at best.

Psychophysiology provides occasional support (e.g. De Pascalis et al., 1996), but it is tellingthat the most comprehensive series of studies testing the theory directly ®nds little consistentsupport for it (Bartussek et al., 1996). In performance paradigms, Gray's theory, likeEysenck's, has di�culties in accommodating variation of personality e�ects with information-processing demands of the task. Motivational variables sometimes have moderating e�ectsbroadly consistent with the theory, although there are concerns about replicability (Hernaiz-Sanders, 1991; Pickering et al., 1997) and discrimination of motivational USs and CSs (cf.Derryberry and Reed, 1994). Gray's theory of Anx has similar problems to Eysenck's theory ofN, including lack of support from psychophysiology (Fahrenberg, 1987), lack of clear-cute�ects of Anx in instrumental learning studies and di�culties in accounting for the variousinformation-processing factors which moderate Anx e�ects in performance studies (M. W.Eysenck, 1992). The problems of anxiety studies are disappointing in view of the impressiveevidence for the BIS as the basis for anxiolytic drug e�ects in animals (Gray, 1982;McNaughton, 1997). It is tempting to conclude that human anxiety is simply much moresusceptible to cognitive control than to the BIS: self-regulative processes are central to humananxiety (Wells and Matthews, 1994), but may lack close analogues in the rat.The Gray theory receives its strongest support from the role of motivational signals as an

important moderator of personality e�ects and from the broad association between E/Imp and``behavioural activation''. Overall, however, our impression is that the theory has had lesspredictive success than the Eysenck theory, even within those domains to which it most clearlyapplies. Pickering et al. (1997, p. 63) admit that di�erent post hoc explanations are needed to ®tthe theory to results from di�erent studies. It is premature to dismiss the theory at this stageand some speci®c paradigms provide quali®ed support for it. Recent commentary on thetheory (Pickering, 1997; Pickering et al., 1997) demonstrates conceptual di�culties in linkingthe BIS and BAS to behavioural measures, in predicting their interaction and in characterizingthe role of arousal. There seem to be di�erences in opinion between Gray's coworkerson development of the model: Pickering (1997) makes arousal the proximal in¯uence onbehaviour, whereas McNaughton's (1997) account of anxiety is based on hippocampalcomputation of threat. The domain of application of Gray's theory is also uncertain: thereseems to have been a retreat from the wide-ranging account given by Gray (1981) to a searchfor a small number of speci®c paradigms which might support the theory (Pickering et al.,1997). Resolution of these di�culties would facilitate theory testing.Tentatively, the evidence reviewed suggests E may have two clusters of psychophysiological

and behavioural correlates. Table 2 distinguishes (1) a ``cortico±reticular'' E, related to classicalarousal indices, conditioning and sensory thresholds and (2) a ``dopaminergic E'' associatedwith motor responsivity, subjective energy, multiple-task performance (controlled by Posner'sevent detection system) and reward sensitivity. The dopaminergic basis for modulation ofmotor activity, positive emotion and reward-incentive behaviour in particular seems quite well-established (Depue, 1995). In addition, molecular genetic data show relationships betweenextraversion and a polymorphism for a dopamine receptor gene (Ebstein et al., 1996). In a tidy


world, these two sets of indices would correlate with distinct primary factors, but the datareviewed provide little basis for doing so1. Contrary to prediction from Gray's theory, Imprelates at least as strongly to factors associated with cortico±reticular E as it does todopaminergic E. Speci®cally, there are studies showing associations between Imp and lowerEEG arousal (O'Gorman and Lloyd, 1987; Stenberg, 1992), reduced phasic EDA in placeboconditions (Smith et al., 1981) and eyelid conditioning (Eysenck and Levey, 1972). Smith etal.'s EDA study (Smith et al., 1981) reported impulsivity� ca�eine interactions suggestive of ahigher threshold for TMI in high impulsives.

7.3. Personality theory and the biological approach

Not so long ago, trait psychologists and social learning theorists such as Mischel (1968)belonged to two opposing, sometimes antagonistic camps, as exempli®ed by the debates of the1960s and 1970s. Trait psychology came with a sharply-de®ned package of assumptions,including the heritability of traits and a causal model assuming feed-forward of individualdi�erences in brain function into behaviour and learning. The resurgence of trait psychologyfollowing on from Eysenck and Eysenck's (1980) decisive refutation of situationism hasprovided considerable evidence in support of these assumptions (Matthews and Deary, 1998).The behaviour and molecular genetic evidence con®rm that biology plays some role inpersonality traits. However, the ``Berlin Wall'' between trait approaches and social learningapproaches is crumbling. Many social learning theorists recognise that traits have at least somevalidity and are exploring social cognitive explanations for their stability and in¯uence onbehaviour (e.g. Snyder, 1992). Conversely, trait psychologists are increasingly developinginformation-processing models within which cognitive factors are more proximal causalin¯uences on individual di�erences than are biological factors (e.g. Revelle, 1993).

Table 2Two types of correlate of extraversion

``Cortico±reticular'' extraversion ``Dopaminergic'' extraversion

Low cortical arousability decreased motoneuronal excitabilityLow autonomic arousability conditioning to rewardInsensitivity to TMI faster movement timePoor eyelid conditioning multiple channel detection

High sensory threshold subjective energy

1 A reviewer of this paper suggested that, on the basis of the supposed dopaminergic basis for psychoticism

(Eysenck, 1997), ``dopaminergic E'' could equally well be regarded as ``dopaminergic P''. However, P does not seemto be related to the criteria we have used to de®ne dopaminergic E, such as positive emotion (Matthews et al., inpress), motoneural excitability (Pivik et al., 1988) and movement time (Doucet and Stelmack, 1997). Corr et al.

(1995a) failed to show any signi®cant correlations between P and conditioning to reward stimuli. Conversely, Prelates to reticulo±cortical indices such as EEG measures (Matthews and Amelang, 1993) and eyelid conditioning(Frcka and Martin, 1987).


In this new era of research, the central issues concern not the basic validity of traits, but thenature of the explanatory constructs to be used in linking traits to individual di�erences inbehaviour. Both the Eysenck and the Gray theories make a clear statement that neuralprocesses are the preferred explanatory construct and the Eysenck theory in particular has hadnotable successes in explaining individual di�erences in responses closely tied to brain function.However, the di�culties which both theories have experienced in explaining the humanperformance data challenge the centrality of neural explanations. Furthermore, di�culty inspecifying moderating factors leads to considerable predictive ambiguity. We can predict thateither extraverts or introverts are more aroused, depending on assumptions made about TMI.Similarly, Gray's theory can predict that Anx is associated with either more or less motorresponsivity, depending on whether behavioural inhibition or arousal e�ects predominate. Thescope for post hoc rationalisation of results is obvious.Thus, trait psychology has two central challenges to meet. First, should we retain individual

di�erences in the functioning of broad neural systems as the principal explanatory construct?Cognitive and social-cognitive models may provide viable alternatives, with more precisespeci®cation of moderating factors. Second, both Eysenck and Gray assume that each traitcorresponds to a single, key underlying system, but there may not be any simple one-to-onemappings between brain functions and traits (Zuckerman, 1991). The twin challenges posed bycognitive theories and by multiple-systems models of traits may be met in various ways. Thebest hope for research in the Eysenck±Gray tradition is methodological improvements inassessment of brain function. Perhaps current psychophysiological techniques simply do notpick up the individual di�erences in neural processes which drive behaviour. A secondapproach, retaining the centrality of neural explanations, is to develop more complexphysiological models (e.g. Zuckerman, 1991), at the possible cost of loss of theoreticalcoherence and testability. Within such an approach, both the Eysenck and Gray theories mayhave partial validity. Perhaps E is independently related to cortico±reticular arousability and toa reward system. A third alternative (Revelle, 1993), is to extend the causal chain linking traitsto behaviour by introducing cognitive variables as additional mediating constructs, so thatneural processes become more distal. The ®nal possibility is to argue that the associationsbetween traits and broad neural systems are simply not strong enough to explain thebehavioural correlates of traits. Matthews (Matthews and Dorn, 1995; Matthews, 1997) hasargued that traits are associated with packages of somewhat independent neural and cognitivefunctions which together support adaptations to specialised social and physical environments.Cognitive constructs may be more appropriate than biological ones for explaining the majorityof behaviours, so that explanations of the kind o�ered by the Eysenck and Gray theories arerelevant to a restricted range of phenomena only. It remains to be seen whether either theoryprovides a comprehensive explanation for the correlates of traits, but, in any case, personalitypsychology owes an enormous debt to both scientists.

Acknowledgements

This article was written at the invitation of Hans Eysenck. We were greatly saddened by hisdeath while the article was in preparation. We thank Je�rey Gray, Alan Pickering and Robert


Stelmack for helpful comments on a previous draft of this paper. We are also grateful to NeilMcNaughton for insights into animal studies of anxiety and to Ana Adan for making availabledata on personality and mood.

References

Adan, A., & GuaÁ rdia, J. (1997).-E�ects of time of day and personality on self-reported activation. Psicothema, 9, 133±143.

Amelang, M., & Ullwer, U. (1991). Correlations between psychometric measures and psychophysiological as well as experimental vari-

ables in studies on extraversion and neuroticism. In J. Strelau & A. Angleitner (Eds.), Explorations in temperament. New York:

Plenum.

Avila, C. (1995).-Facilitation and inhibition of visual orienting as a function of personality. Personality and Individual Di�erences, 18,

503±510.

Barrett, E. S. (1971).-Psychophysiological correlates of classical di�erential eyelid conditioning among subjects selected on the basis of

impulsiveness and anxiety. Biological Psychiatry, 3, 339±346.

Bartussek, D. (1996). Extraversion and incidental verbal learning: An experimental decision between predictions from Eysenck's and

Gray's theories of extraversion. Paper presented to the 8th European Conference on Personality, Ghent.

Bartussek, D., Becker, G., Diedrich, O., Naumann, E., & Maier, S. (1996).-Extraversion, neuroticism and event-related potentials in

response to emotional stimuli. Personality and Individual Di�erences, 20, 301±312.

Bartussek, D., Diedrich, O., Naumann, E., & Collet, W. (1993).-Introversion±extraversion and event related potential (ERP): A test of

J. A. Gray's theory. Personality and Individual Di�erences, 14, 565±574.

Bartussek, D., Naumann, E., Collet, W., & Moeller, H. (1990a).-Ereigniskorreliertes Hirnrindenpotential (EKP) in Abhangigkeit von

der Rueckmeldehedeutung der Reize, Extraversion und Neurotizismus. Trierer Psychologische Berichte, 17, 1±56.

Bartussek, D., Naumann, E., Moeller, H., Vogelbacher, D., & Diedrich, O. (1990b). Emotionale Wortbedeutung, Extraversion, und

der spaete positive Komplex des Ereigniskorrelieerten Potentials (EKP). Lecture given by O. Diedrich at the 37th Congress of the

German Psychological Society in Kiel.

Beatty, J. (1986). Computation, control and energetics: A biological perspective. In G. R. J. Hockey, A. W. K. Gaillard, & M. G. H.

Coles (Eds.), Energetics and human information processing. Dordrecht: Martinus Nijho�.

Blackburn, I. M., Cameron, C. M., & Deary, I. J. (1990).- Individual di�erences and response to the Velten Mood Induction

Procedure. Personality and Individual Di�erences, 11, 725±731.

Bolger, E. A., & Schilling, E. A. (1991).-Personality and the problems of everyday life: The role of neuroticism in exposure and reac-

tivity to daily stressors. Journal of Personality, 59, 335±386.

BrandstaÈ tter, H. (1994).-Well-being and motivated person-environment ®t: A time-sampling study of emotions. European Journal of

Personality, 8, 75±94.

Brebner, J., & Cooper, C. (1985). A proposed uni®ed model of extraversion. In J. T. Spence & C. E. Izard (Eds.), Motivation, emotion

and personality. Amsterdam: North-Holland.

Brocke, B., Tasche, K. G., & Beauducel, A. (1997).-Biopsychological foundations of extraversion: Di�erential e�ort reactivity and

state contol. Personality and Individual Di�erences, 22, 447±458.

Bullock, W. A., & Gilliland, K. (1993).-Eysenck's arousal theory of introversion±extraversion: A converging measures investigation.

Journal of Personality and Social Psychology, 64, 113±123.

Busch, H. A., King, P. R., & Guttman, M. A. (1994).-Interactional anxiety in an examination situation with French and English high

school students: An empirical test of a composite predictor for state anxiety. Personality and Individual Di�erences, 17, 111±116.

Church, M. W., & Shucard, D. W. (1987).- Theophylline-induced changes in the mouse brainstem auditory evoked potential.

Neurotoxicology and Teratology, 9, 59±66.

Coles, M. G. H., Gratton, G., & Fabiani, M. (1990). Event-related potentials. In J. T. Cacioppo & L. G. Tassinary (Eds.), Principles of

psychophysiology: Physical, social and inferential elements. Cambridge: Cambridge University Press.

Corr, P. J., Kumari, V., Wilson, G. D., Checkley, S., & Gray, J. A. (1997).-Harm avoidance and a�ective modulation of the startle

re¯ex: A replication. Personality and Individual Di�erences, 22, 591±593.

Corr, P. J., Pickering, A. D., & Gray, J. A. (1995a).- Personality and reinforcement in associative and instrumental learning.

Personality and Individual Di�erences, 19, 47±72.

Corr, P. J., Pickering, A. D., & Gray, J. A. (1995b).-Sociability/impulsivity and ca�eine-induced arousal: Critical/¯icker fusion fre-

quency and procedural learning. Personality and Individual Di�erences, 18, 713±730.

Costa, P. T., & McCrae, R. R. (1980).-In¯uence of extraversion and neuroticism on subjective well-being: Happy and unhappy

people. Journal of Personality and Social Psychology, 38, 668±678.

Davies, D. R., & Parasuraman, R. (1982). The psychology of vigilance. London: Academic Press.


Dawson, M. E., Schell, A. M., & Filion, D. L. (1990). The electrodermal system. In J. T. Cacioppo & L. G. Tassinary (Eds.), Principles

of psychophysiology: Physical, social and inferential elements. Cambridge: Cambridge University Press.

De Pascalis, V., Fiore, A. D., & Sparita, A. (1996).-Personality, event-related potential (ERP) and heart rate (HR): An investigation of

Gray's theory. Personality and Individual Di�erences, 20, 733±746.

Depue, R. A. (1995).-Neurobiological factors in personality and depression. European Journal of Personality, 9, 413±439.

Derryberry, D. (1987).-Incentive and feedback e�ects on target detection: A chronometric analysis of Gray's model of temperament.


Derryberry, D., & Reed, M. A. (1994).-Temperament and attention: Orienting toward and away from positive and negative signals.


Derryberry, D., & Reed, M. A. (1997). Motivational and attentional components of personality. In G. Matthews (Ed.), Cognitive

science perspectives on personality and emotion. Amsterdam: Elsevier Science B.V.

Dickman, S. J., & Meyer, D. E. (1988).-Impulsivity and speed-accuracy tradeo�s in information processing. Journal of Personality and

Social Psychology, 54, 274±290.

Di Traglia, G., & Polich, J. (1991).-P300 and introverted/extraverted personality types. Psychophysiology, 28, 177±184.

Dorn, L., & Matthews, G. (1995).-Prediction of mood and risk appraisals from trait measures: Two studies of simulated driving.

European Journal of Personality, 9, 25±42.

Doucet, C., & Stelmack, R. M. (1997).-Movement time di�erentiates extraverts from introverts. Personality and Individual Di�erences,

23, 775±786.

Ebstein, R. P., Novick, O., Umansky, R., Priel, B., Osher, Y., Blaine, D., Bennett, E. R., Nemanov, L., Katz, M., & Belmaker, R. H.

(1996).-Dopamine D4 receptor (D4DR) exon III polymorphism associated with the human personality trait of novelty seeking.

Nature Genetics, 12, 78±80.

Emmons, R. A., & Diener, E. (1986).-In¯uence of impulsivity and sociability on subjective well-being. Journal of Personality and Social

Psychology, 50, 1211±1215.

Endler, N. S., Parker, J. D. A., Bagby, R. M., & Cox, B. J. (1991).-The multidimensionality of state and trait anxiety: The factor

structure of the Endler Multidimensional Anxiety Scales. Journal of Personality and Social Psychology, 60, 919±926.

Eysenck, H. J. (1967). The biological basis of personality. Spring®eld: Thomas.

Eysenck, H. J. (1981). General features of the model. In H. J. Eysenck (Ed.), A model for personality. Berlin: Springer-Verlag.

Eysenck, H. J. (1992).-The de®nition and measurement of psychoticism. Personality and Individual Di�erences, 13, 757±786.

Eysenck, H. J. (1994). Personality: Biological foundations. In P. A. Vernon (Ed.), The neuropsychology of individual di�erences.

London: Academic Press.

Eysenck, H. J. (1997).-Personality and experimental psychology: The uni®cation of psychology and the possibility of a paradigm.


Eysenck, H. J., & Eysenck, M. W. (1985). Personality and individual di�erences. New York: Plenum.

Eysenck, H. J., & Levey, A. (1972) Conditioning, extraversion-introversion and the strength of the nervous system. In V. D. Nebylitsyn

and J. A. Gray (Eds.), Biological bases of individual behaviour. London: Academic Press.

Eysenck, M. W. (1981) Learning, memory and personality. In H. J. Eysenck (Ed.), A model for personality. Berlin: Springer-Verlag.

Eysenck, M. W. (1982) Attention and arousal: Cognition and performance. New York: Springer-Verlag.

Eysenck, M. W. (1992) Anxiety: The cognitive perspective. Hillsdale, NJ: Erlbaum.

Eysenck, M. W., & Eysenck, H. J. (1980).-Mischel and the concept of personality. British Journal of Psychology, 71, 191±204.

Eysenck, M. W., & Eysenck, M. C. (1979).-Memory scanning, introversion±extraversion and levels of processing. Journal of Research

in Personality, 13, 305±315.

Eysenck, M. W., & Hepburn, L. (1989).-Personality, average mood and mood variability. Personality and Individual Di�erences, 10,

975±983.

Fahrenberg, J. (1987). Concepts of activation and arousal in the theory of emotionality (neuroticism): A multivariate conception. In J.

Strelau & H. J. Eysenck (Eds.), Personality dimensions and arousal. New York: Plenum.

Fowles, D. C. (1980).-The three arousal model: Implications of Gray's two-factor learning theory for heart rate, electrodermal activity

and psychopathy. Psychophysiology, 17, 87±104.

Fowles, D. C., Roberts, R., & Nagel, K. (1977).-The in¯uence of introversion±extraversion on the skin conductance response of stress

and stimulus intensity. Journal of Research in Personality, 11, 129±146.

Frcka, G., Beyts, J., Levey, A. B., & Martin, I. (1983).-The in¯uence of psychoticism on classical conditioning. Personality and

Individual Di�erences, 4, 189±197.

Frcka, G., & Martin, I. (1987).-Is there Ð or is there not Ð an in¯uence of impulsiveness on classical eyelid conditioning?. Personality

and Individual Di�erences, 8, 241±252.

Furnham, A. (1981).-Personality and activity preference. British Journal of Social Psychology, 20, 57±60.

Gale, A. (1973). The psychophysiology of individual di�erences: Studies of extraversion and the EEG. In P. Kline (Ed.), New

approaches in psychological measurement. New York: Wiley.


Gale, A. (1983).-Electroencephalographic studies of extraversion-introversion: A case study in the psychophysiology of individual

di�erences. Personality and Individual Di�erences, 4, 371±380.

Gale, A. (1984).-O'Gorman vs. Gale: A reply. Biological Psychology, 19, 129±136.

Geen, R. G. (1983). The psychophysiology of extraversion±introversion. In J. T. Cacioppo & R. E. Petty (Eds.), Social psychophysiol-

ogy. New York: Guilford.

Geen, R. G. (1985).-Test anxiety and visual vigilance. Journal of Personality and Social Psychology, 49, 963±970.

Giambra, L. M., Quilter, R. E., & Phillips, P. B. (1989).-The relationship of age and extraversion to arousal and performance on a

sustained attention task: A cross-sectional investigation using the Mackworth clock-test. Personality and Individual Di�erences, 9,

225±230.

Golombok, S., Stavrou, A., & Bonn, J. (1991).-The e�ects of diazepam on anxiety-related cognition. Cognitive Therapy and Research,

15, 459±467.

Gray, J. A. (1970).-The psychophysiological basis of introversion±extraversion. Behaviour Research and Therapy, 8, 249±266.

Gray, J. A. (1981). A critique of Eysenck's theory of personality. In H. J. Eysenck (Ed.), A model for personality. Berlin: Springer-

Verlag.

Gray, J. A. (1982) The neuropsychology of anxiety: An enquiry into the functions of the septo-hippocampal system. Oxford: Oxford

University Press.

Gray, J. A. (1987a). The psychology of fear and stress (2nd ed.). Cambridge: Cambridge University Press.

Gray, J. A. (1987b).-Perspectives on anxiety and impulsivity: A commentary. Journal of Research in Personality, 21, 493±509.

Gray, J. A. (1991). Neural systems, emotion and personality. In J. Madden IV (Ed.), Neurobiology of learning, emotion and a�ect. New

York: Raven Press.

Gray, J. A., Feldon, J., Rawlins, J., Helmsley, D., & Smith, A. (1991).-The neuropsychology of schizophrenia. Behavioural and Brain

Sciences, 14, 1±84.

Gupta, B. S., & Gupta, U. (1984).-Dextroamphetamine and individual susceptibility to reinforcement in verbal and operant condition-

ing. British Journal of Psychology, 75, 201±208.

Gupta, B. S., & Shukla, A. P. (1989).-Verbal operant conditioning as a function of extraversion and reinforcement. British Journal of

Psychology, 80, 39±44.

Gupta, S. (1990).-Impulsivity/sociability and reinforcement in verbal operant conditioning: A replication. Personality and Individual

Di�erences, 11, 585±590.

Hernaiz-Sanders, H. (1991) Human reactions to arousal and reinforcement: Implications for the relationship between Eysenck's and

Gray's models of personality. Unpublished doctoral dissertation, University of London.

Hockey, G. R. J. (1984). Varieties of attentional state: The e�ects of the environment. In R. Parasuraman & D. R. Davies (Eds.),

Varieties of attention. New York: Academic Press.

Howarth, E., & Eysenck, H. J. (1968).- Extraversion, arousal and paired-associate recall. Journal of Experimental Research in

Personality, 3, 114±116.

Jewett, D. L., Romano, M. N., &Williston, J. S. (1970).-Human auditory evoked potentials: Possible brainstem components detected

on the scalp. Science, 167, 1517±1518.

Koelega, H. S. (1992).-Extraversion and vigilance performance: 30 years of inconsistencies. Psychological Bulletin, 112, 239±258.

Kumari, V., Corr, P. J., Wilson, G. D., Kaviani, H., Thornton, J. C., Checkley, S. A., & Gray, J. A. (1996).-Personality and modu-

lation of the startle re¯ex by emotionally-toned ®lmclips. Personality and Individual Di�erences, 21, 1029±1041.

Lacey, J. I. (1967) Somatic response patterning and stress: Some revisions of activation theory. In M. H. Appley & R. Trumbull (Eds.),

Psychological stress: Issues in research. New York: Appleton-Century-Crofts.

Larsen, R. J., & Ketelaar, T. (1991).-Personality and susceptibility to positive and negative emotional states. Journal of Personality and


Leon, M. R., & Revelle, W. (1985).-E�ects of anxiety on analogical reasoning: A test of three theoretical models. Journal of Personality

and Social Psychology, 49, 1302±1315.

Levey, A. B., & Martin, I. (1981) Personality and conditioning. In H. J. Eysenck (Ed.), A model for personality. Berlin: Springer-Verlag.

Lukas, J. H. (1980).-Human auditory attention: The olivocochlear bundle may function as a peripheral ®lter. Psychophysiology, 17,

444±452.

MacAndrew, C., & Steele, T. (1991).-Gray's behavioral inhibition system: A psychometric examination. Personality and Individual

Di�erences, 12, 157±171.

Mangan, G. L. (1982) The biology of human conduct. Oxford: Pergamon.

Mangan, G. L., & Hookway, D. (1988).-Perception and recall of aversive material as a function of personality type. Personality and


Matthews, G. (1985).-The e�ects of extraversion and arousal on intelligence test performance. British Journal of Psychology, 76, 479±

493.

Matthews, G. (1988).-Morningness-eveningness as a dimension of personality: trait, state and psychophysiological correlates. European

Journal of Personality, 2, 277±293.


Matthews, G. (1992a) Mood. In A. P. Smith & D. M. Jones (Eds.), Handbook of human performance. Vol. 3: State and trait. London:

Academic.

Matthews, G. (1992b). Extraversion. In A. P. Smith & D. M. Jones (Eds.), Handbook of human performance. Vol. 3: State and trait.

London: Academic.

Matthews, G. (1996).-Signal probability e�ects on high-workload vigilance tasks. Psychonomic Bulletin and Review, 3, 339±343.

Matthews, G. (1997) Extraversion, emotion and performance: A cognitive-adaptive model. In G. Matthews (Ed.), Cognitive science

perspectives on personality and emotion. Amsterdam: Elsevier Science B.V.

Matthews, G., & Amelang, M. (1993).-Extraversion, arousal theory and performance: A study of individual di�erences in the EEG.


Matthews, G., & Davies, D. R. (1998). Arousal and vigilance: The role of task factors. In R. B. Ho�man, M. F. Sherrick, & J. S. Warm

(Eds.), Viewing psychology as a whole: The integrative science of William N. Dember. Washington, DC: American Psychological

Association.

Matthews, G., Davies, D. R., & Lees, J. L. (1990b).-Arousal, extraversion and individual di�erences in resource availability. Journal

of Personality and Social Psychology, 59, 150±168.

Matthews, G., & Deary I. (1998). Personality traits. Cambridge: Cambridge University Press.

Matthews, G., & Dorn, L. (1995). Personality and intelligence: Cognitive and attentional processes. In D. H. Saklofske & M. Zeidner

(Eds.), International handbook of personality and intelligence. New York: Plenum.

Matthews, G., Jones, D. M., & Chamberlain, A. G. (1990a).-Re®ning the measurement of mood: The UWIST Mood Adjective

Checklist. British Journal of Psychology, 81, 17±42.

Matthews, G., Joyner, L., Gilliland, K., Campbell, S. E., Falconer, S., & Huggins, J. (in press). Validation of a comprehensive stress

state questionnaire: Towards a state Big Three? In I. Mervielde, I. J. Deary, F. De Fruyt, & F. Ostendorf (Eds.), Personality psy-

chology in Europe, Vol. 7. Tilburg: Tilburg University Press.

Matthews, G., & Harley, T. A. (1993).-E�ects of extraversion and self-report arousal on semantic priming: A connectionist approach.


McNally, R. J., Riemann, B. C., & Kim, E. (1990).-Selective processing of threat cues in panic disorder. Behaviour Research and

Therapy, 28, 407±412.

McNaughton, N. (1985).- Chlordiapoxide and successive discrimination: Di�erent e�ects on acquisition and performance.

Pharmacology, Biochemistry and Behavior, 23, 487±494.

McNaughton, N. (1997).-Cognitive dysfunction resulting from hippocampal hyperactivity Ð A possible cause of anxiety disorder?.

Pharmacology, Biochemistry and Behavior, 56, 603±611.

Meyer, G. J., & Shack, J. R. (1989).-Structural convergence of mood and personality: Evidence for old and new directions. Journal of

Personality and Social Psychology, 57, 691±706.

Mischel, W. (1968). Personality and assessment. New York: Wiley.

Myrtek, M. (1984). Constitutional psychophysiology. London: Academic Press.

Naveteur, J., & Freixa-i-Baque, E. (1987).-Individual di�erences in electrodermal activity as a function of subject's anxiety. Personality

and Individual Di�erences, 8, 615±626.

Naveteur, J., & Roy, J. C. (1990).-Electrodermal activity of low and high trait anxiety subjects during a frustrative video game. Journal

of Psychophysiology, 4, 221±227.

Nichols, S. L., & Newman, J. P. (1986).-E�ects of punishment on response latency in extraverts. Journal of Personality and Social

Psychology, 50, 624±630.

O'Gorman, J. G. (1977).-Individual di�erences in habituation of human physiological responses: A review of theory, method and ®nd-

ings in the study of personality correlates in non-clinical populations. Biological Psychology, 5, 257±318.

O'Gorman, J. G. (1984).-Extraversion and the EEG: I. An evaluation of Gale's hypothesis. Biological Psychology, 19, 95±112.

O'Gorman, J. G., & Lloyd, J. E. M. (1987).- Extraversion, impulsiveness and EEG alpha activity. Personality and Individual

Di�erences, 8, 169±174.

Orlebeke, J. F., & Feij, J. A. (1979). The orienting re¯ex as a personality correlate. In E. H. Van Olst & J. F. Orlebeke (Eds.), The

orienting re¯ex in humans (pp. 567±585). Hillsdale, NJ: Erlbaum.

Paisey, T. J. H., & Mangan, G. L. (1988).- Personality and conditioning with appetitive and aversive stimuli. Personality and


Panksepp, J. (1982).-Toward a general psychobiological theory of emotion. Behavioural and Brain Sciences, 5, 407±422.

Pearson, G. L., & Freeman, F. G. (1991).-E�ects of extraversion and mental arithmetic on heart rate reactivity. Perceptual and Motor

Skills, 72, 1235±1244.

Pickering, A. D. (1997).-The conceptual nervous system and personality: From Pavlov to neural networks. European Psychologist, 2,

139±163.

Pickering, A. D., Corr, P. J., Powell, J. H., Kumari, V., Thornton, J. C., & Gray, J. A. (1997). Individual di�erences in reactions to

reinforcing stimuli are neither black nor white: To what extent are they Gray? In H. Nyborg (Ed.), The scienti®c study of human

nature: Tribute to Hans J. Eysenck at eighty. Amsterdam: Elsevier Science B.V.


Pickering, A. D., DõÂ az, A., & Gray, J. A. (1995).- Personality and reinforcement: An exploration using a maze-learning task.


Pivik, R. T., Stelmack, R. M., & Bylsma, F. W. (1988).-Personality and individual di�erences in spinal motoneuronal excitability.

Psychophysiology, 25, 16±24.

Posner, M. I., & Raichle, M. E. (1994). Images of mind. New York: Scienti®c American Library.

Posner, M. I., & Rothbart, M. K. (1991). Attentional mechanisms and conscious experience. In D. Milner & M. Rugg (Eds.), The

neuropsychology of consciousness. San Diego, CA: Academic Press.

Revelle, W. (1993). Individual di�erences in personality and motivation: ``Non-cognitive'' determinants of cognitive performance. In

A. Baddeley & L. Weiskrantz (Eds.), Attention: Selection, awareness and control. Oxford: Oxford University Press.

Revelle, W., & Loftus, D. A. (1992). The implications of arousal e�ects for the study of a�ect and memory. In S.-AÊ . Christianson (Ed.),

The handbook of emotion and memory: Research and theory. Hillsdale, NJ: Erlbaum.

Revelle, W., Humphreys, M. S., Simon, L., & Gilliland, K. (1980).-The interactive e�ect of personality, time of day and ca�eine: A test

of the arousal model. Journal of Experimental Psychology: General, 109, 1±31.

Richards, M., & Eves, F. (1991).-Personality, temperament and the cardiac defense response. Personality and Individual Di�erences, 12,

999±1004.

Robbins, T. W. (1986). Psychopharmacological and neurobiological aspects of the energetics of information processing. In G. R. J.

Hockey, A. W. K. Gaillard, & M. G. H. Coles (Eds.), Energetics and human information processing. Dordrecht: Martinus Nijho�.

Rusting, C. L., & Larsen, R. J. (1995).-Moods as sources of stimulation: Relationships between personality and desired mood states.


Rusting, C. L., & Larsen, R. J. (1997).-Extraversion, neuroticism and susceptibility to positive and negative a�ect: A test of two theor-

etical models. Personality and Individual Di�erences, 22, 607±612.

Shigehisa, T., & Symons, J. R. (1973).-E�ects of intensity of visual stimulation on auditory sensitivity in relation to personality. British

Journal of Psychology, 64, 205±213.

Shigehisa, T., Shigehisa, P. M. J., & Symons, J. R. (1973).-E�ect of interval between auditory and preceding visual stimuli on audi-

tory sensitivity. British Journal of Psychology, 64, 367±373.

Smith, B. D. (1983).-Extraversion and electrodermal activity: Arousability and the inverted U. Personality and Individual Di�erences,

4, 411±419.

Smith, B. D., Kline, R., Lindgren, K., Ferro, M., Smith, D. A., & Nespor, A. (1995).-The lateralized processing of a�ect in emotion-

ally labile extraverts and introverts: central and autonomic e�ects. Biological Psychology, 39, 143±157.

Smith, B. D., Rypma, C. B., & Wilson, R. J. (1981).-Dishabituation and spontaneous recovery of the electrodermal orienting re-

sponse: E�ects of extraversion, impulsivity, sociability and ca�eine. Journal of Research in Personality, 15, 233±240.

Snyder, M. (1992). Motivational foundations of behavioral con®rmation. In M. P. Zanna (Ed.), Advances in experimental social psy-

chology, Vol. 25. New York: Academic Press.

Spence, K. W. (1964).-Anxiety (drive) level and performance in eyelid conditioning. Psychological Bulletin, 61, 129±139.

Stelmack, R. M. (1981). The psychophysiology of extraversion and neuroticism. In H. J. Eysenck (Ed.), A model for personality. Berlin:

Springer-Verlag.

Stelmack, R. M. (1990).-Biological bases of extraversion: Psychophysiological evidence. Journal of Personality, 58, 293±311.

Stelmack, R. M., Campbell, K. B., & Bell, I. (1993a).-Extraversion and brainstem auditory evoked potentials during sleep and wake-

fulness. Personality and Individual Di�erences, 14, 447±453.

Stelmack, R. M., & Houlihan, M. (1995) Event-related potentials, personality and intelligence: Concepts, issues and evidence. In D. H.

Saklofske & M. Zeidner (Eds.), International handbook of personality and intelligence. New York: Plenum.

Stelmack, R. M., Houlihan, M., & McGarry Roberts, P. A. (1993b).-Personality, reaction time and event-related potentials. Journal

of Personality and Social Psychology, 65, 399±409.

Stelmack, R. M., & Wilson, K. (1982).-Extraversion and the e�ects of frequency and intensity on the auditory brainstem evoked re-

sponse. Personality and Individual Di�erences, 3, 373±380.

Stenberg, G. (1992).-Personality and the EEG: Arousal and emotional arousability. Personality and Individual Di�erences, 13, 1097±

1113.

Stenberg, G. (1994).-Extraversion and the P300 in a visual classi®cation task. Personality and Individual Di�erences, 16, 543±560.

Stenberg, G., Risberg, J., Warkentin, S., & Rosen, I. (1990).-Regional patterns of cortical blood ¯ow distinguishing extraverts from

introverts. Personality and Individual Di�erences, 11, 663±673.

Swickert, R. J. (1996). Relationship between the brainstem auditory evoked response and extraversion, impulsivity and sociability.

Doctoral dissertation, University of Oklahoma, 1990. Dissertation Abstracts International, 57/05, 3446.

Tankova, I., Adan, A., & Buela-Casal, G. (1994).-Circadian typology and individual di�erences: A review. Personality and Individual

Di�erences, 16, 671±684.

Tecce, J. J. (1972).-Contingent negative variation and psychological processes in man. Psychological Bulletin, 77, 73±108.

Thayer, R. E. (1989). The biopsychology of mood and arousal. Oxford: Oxford University Press.

Thayer, R. E. (1996). The origin of everyday moods. New York: Oxford University Press.


Thayer, R. E., Newman, J. R., & McClain, T. M. (1994).-Self-regulation of mood: Strategies for changing a bad mood, raising energy

and reducing tension. Journal of Personality and Social Psychology, 67, 910±925.

Wallace, J. F., & Newman, J. P. (1990).-Di�erential e�ects of reward and punishment cues on response speed in anxious and impul-

sive individuals. Personality and Individual Di�erences, 11, 999±1009.

Wallace, J. F., Newman, J. P., & Bachorowski, J.-A. (1991).-Failures of response modulation: Impulsive behaviour in anxious and

impulsive individuals. Journal of Research in Personality, 25, 23±44.

Watson, D., & Clark, L. A. (1992).-On traits and temperament: General and speci®c factors of emotional experience and their relation

to the ®ve-factor model. Journal of Personality, 60, 441±476.

Watson, D., Clark, L. A., McIntyre, C. W., & Hamaker, S. (1992).-A�ect, personality and social activity. Journal of Personality and


Wells, A., & Matthews, G. (1994). Attention and emotion: A clinical perspective. Hove: Erlbaum.

Werre, P. F. (1986). Contingent negative variation: Relation to personality and modi®cation by stimulation of sedation. In J. Strelau,

F. Farley, & A. Gale (Eds.), The biological basis of personality and behavior, Vol. 2. London: Hemisphere.

Wigglesworth, M. J., & Smith, B. D. (1976).-Habituation and dishabituation of the electrodermal orienting re¯ex in relation to extra-

version and neuroticism. Journal of Research in Personality, 10, 437±445.

Williams, D. G. (1989).-Personality e�ects in current mood: Pervasive or reactive?. Personality and Individual Di�erences, 10, 941±948.

Wilson, G. D. (1990).-Personality, time of day and arousal. Personality and Individual Di�erences, 11, 153±168.

Wilson, G. D., Barrett, P. T., & Gray, J. A. (1989).-Human reactions to reward and punishment: A questionnaire examination of

Gray's personality theory. British Journal of Psychology, 80, 509±515.

Zinbarg, R., & Revelle, W. (1989).-Personality and conditioning: A test of four models. Personality and Social Psychology, 57, 301±314.

Zuckerman, M. (1991). Psychobiology of personality. Cambridge: Cambridge University Press.


The personality theories of H. J. Eysenck and J. A. Gray: a comparative review

Documents