Journal o f Personality and Social Psychology
1989, Vol. 56, No. 4, 6 2 9 - 6 3 9
Copyright 1989 by the American Psychological Association, Inc.
0022-3514/89/$00.75
Interactive Effects of Extraversion and Arousal on Attentional Task
Performance: Multiple Resources or Encoding Processes?
Gerald Matthews
University of Aston
Birmingham, England
Dylan M. Jones and A. Graham Chamberlain
University of Wales Institute of Science and Technology
Cardiff, Wales
Reports study of the effects on 6 tasks of time of day, extraversion, and self-report arousal. We
selected tasks to test processing resource explanations of interactive effects of extraversion and
arousal on performance. Significant interactive effects of time of day, extraversion, and arousal on
performance consistent with previous research were found with 2 tasks. One of these tasks was
attentional (speed of serial reaction), the other required short-term memory (digit span). The interactions we found do not support the common assumption that effects of extraversion on performance
are mediated by arousal. Also, the task specificity of the Extraversion × Arousal interactions is
inconsistent with existing processing resource theories of personality. An alternative hypothesis,
which resolves these difficulties, is that time of day, extraversion, and arousal interactively affect
stimulus encoding at an early stage of analysis.
A basic finding in personality is the sensitivity of task performance to interactive effects of extraversion and stressors. Typically, extraverts perform better under arousing conditions, but
introverts' performance is superior under low arousal (Corcoran, 1965)• H. J. Eysenck (1967) has explained such interactions in terms of the inverted-I.I curve said to relate arousal to
performance (Duffy, 1957). Under most conditions, extraversion is negatively correlated with cortical arousal, so that arousing stressors raise arousal to the optimum for performance in
extraverts but cause overarousal in introverts.
Revelle, Humphreys, Simon, and GiUiland (1980) reported.
data inconsistent with the H. J. Eysenck (1967) theory. Their
studies showed, first, that interactive effects of extraversion and
the stimulant drug caffeine on verbal ability test performance
were mediated by the impulsivity rather than the sociability
subfactor of extraversion. Second, effects of impulsivity on performance were critically dependent on time of day. In the morning, high impulsives performed better under caffeine, as expected. In the evening, however, caffeine improved the performance of low impulsives but impaired that of high impulsives,
contrary to expectation. Revelle et al. claimed that the relation
between impulsivity and arousal is dependent on time of day:
Low impulsives are more aroused in the morning, high impulsives in the evening. Humphreys and Revelle (1984) elaborated
this hypothesis within a cognitive model. In this model, the relation between arousal and task performance depends on the ex-
tent to which the task draws on two pools of processing resources: sustained information processing (SIT) and short-term
memory (STM). High arousal increases the availability of SIT
resources and thus tends to improve performance of tasks such
as letter cancellation. High arousal also decreases STM resources and impairs performance of STM tasks. Complex tasks
such as ability tests are said to require both kinds of resource,
and an inverted-I.I relation between arousal and performance
can be derived for such tasks.
Testing T h e o r i e s o f Effects o f P e r s o n a l i t y
on P e r f o r m a n c e
The causal chain posited by Humphreys and Revelle (1984)
is that impulsivity and time of day interactively affect arousal,
which in turn affects resources, which will affect performance if
•
/.
.
.
the task Is appropriately resource-hmlted. H. J. Eysenck (1967)
proposed a similar but simpler causal chain. Thus, both the
Eysenck and the Humphreys and Revelle models see the effects
of extraversion or impulsivity as being arousal-mediated. This
causal sequence can be tested by measuring arousal directly,
allowing independent tests of (a) the relation between personality and arousal and (b) that between arousal and performance.
Arousal can be most readily measured by self-report. Such
measures correlate with composite measures of autonomic
arousal and provide better indicators of broad, integrated
arousal systems than do individual psychophysiological measures (Matthews, 1987; Thayer, 1978). Thayer (1978, 1986) distinguished two dimensions of arousal, energetic and tense
arousal, and showed that it is energetic rather than tense arousal
that predicts cognitive functioning. Energetic arousal is sensitive to motor activity; tense arousal is associated with stress and
emotional reactions. On these grounds, Thayer suggested that
We are grateful to two anonymous referees for comments on a previous draft of this article.
Correslxmdence concerning this article should be addressed to Gerald Matthews, Division of Applied Psychology,University of Aston, Aston Triangle, Birmingham B4 7ET, England.
629
630
G. MATTHEWS, D. JONES, AND G. CHAMBERLAIN
energetic arousal is similar to H. J. Eysenck's (1967) corticoreticular arousal, tense arousal to Eysenck's visceral brain
arousal. Energetic arousal is therefore the preferred measure for
studies of arousal and performance. Several experiments have
suggested that extraversion and measured arousal or activation
interact much as do extraversion and arousal manipulations
such as noise and caffeine. M. W. Eysenck (1981) has shown
that extraverts' retrieval from memory is generally faster under
high activation, whereas that of introverts is faster under low
activation (subject to some modification by task factors). Matthews (1985) found an interactive effect of time of day, extraversion, and activation on intelligence test performance equivalent
to that found by Revelle et al. (1980) with impulsivity and
caffeine. One difficulty with relating self-report arousal measures to cortical arousal is a lack of research on correlations
between self-report arousal and the EEG, although arousing
agents such as caffeine do appear to affect both self-report and
EEG measures (Clubley, Bye, Henson, Peck, & Riddington,
1979).
Two studies in the present program of research have suggested that simple attentional tasks are sensitive to interactive
effects of extraversion and self-report arousal. Matthews (in
press) investigated performance of a self-paced sustained attention task requiring detection of digits. Errors in changing the
target digit according to a simple rule could be distinguished
from simple detection errors. The former are related to efficiency of an "upper" executive level of processing, the latter to
a "lower" stimulus-driven level of processing (Jones & Coyle,
1983). Significant interactive effects of time of day, extraversion,
and arousal on lower-level errors (see Figure 1) were similar to
Matthews's (1985) intelligence test data. There were no significant effects of independent variables on upper-level errors. Matthews and Chappelow (1986), using a selective attention task in
the morning only, also found the characteristic Extraversion X
Arousal interaction, irrespective of selection cue and distractor
presence or absence. These two studies suggested that extraversion and arousal affect efficiency of stimulus detection but not
more complex processes of executive control and selection of
information.
Another important feature of both the Matthews (1985, in
press; Matthews & Chappelow, 1986) studies and the research
of M. W. Eysenck (1977, p. 201) is that interactions between
ext?aversion and arousal were not contingent on any significant
relation between the two variables. In other words, extraversion
and time of day appear to affect the regression of arousal on
performance (Matthews, 1985) rather than changing the positions of subject groups along a single arousal-performance
curve.
Aims and Hypotheses
The primary aims of this study were to test relations between
time of day, extraversion, and arousal that were predicted by the
work of H. J. Eysenck (1967), Humphreys and Revelle (1984),
and Matthews (1985, in press).
We selected tasks as follows. We considered two tasks to be
demanding only of SIT resources: five-choice serial reaction and
a vigilance task. Five-choice serial reaction is sensitive to effects
of a variety of stressors (Broadbent, 197 t). The vigilance task
requires subjects to detect visually degraded stimuli, while placing a minimal load on memory. Parasuraman (1985) has used
a secondary probe stimulus technique to demonstrate that this
kind of task is resource-limited. We also used two memory
tasks: digit span and the Posner, Boies, Eichelman, and Taylor
(1969) letter-match task. The role of attentional resources in
STM is controversial. However, research on working memory
(Baddeley, 1986) has shown interference between concurrent
digit retention and other memory tasks attributable to a "central executive," suggesting that standard digit span tasks affect
resources common to a range of STM tasks. A more finegrained approach to resources in STM can be derived from the
Posner et al. letter-match task, which provides a measure of
speed of access to letter codes, presumably one of the component processes in short-term retention of letters. Secondary
probe task experiments show that this process is resource-limited when subjects are unpracticed on the task (Paap & Ogden,
1981). We used one task considered to require both kinds of
resources: a search for 6 letters within a 20-letter string. The
final measure used was a control for individual differences in
purely motor processes: speed of repeated tapping.
Given these tasks, the following predictions can be made. The
H. J. Eysenck (1967) theory predicted (a) a negative correlation
between extraversion and arousal at both times of day and (b)
regressions of performance on arousal consistent with the
Yerkes-Dodson law. Arousal should be related to performance
positively for easy tasks, negatively for difficult tasks, and by an
inverted U for tasks of intermediate difficulty.
The Humphreys and Revelle (1984) theory predicted (a) an
interactive effect of time of day and extraversion-impulsivity on
arousal and (b) task-dependent effects of arousal on performance. Arousal should facilitate performance on the vigilance
and serial reaction tasks, should impair digit span and retrieval
of letter codes, and should be related to six-letter search by an
inverted U. An absence of effects of arousal on performance
might in principle be explained by claiming that the task was
in fact data- rather than resource-limited.
Predictions can be derived from Matthews's studies (1985,
1987, in press) as follows, using Thayer's (1978) energetic
arousal dimensions as a measure of cortical arousal. Matthews's
(1987) review of studies of extraversion and self-report arousal
predicted a weak positive correlation between extraversion and
energetic arousal. However, the main prediction was that time
of day, extraversion, and measured arousal should affect performance interactively, as in the Matthews (1985) study. If the
mechanism is one of stimulus detection, then strongest effects
should be obtained with the purely attentional tasks used in this
study, serial reaction and vigilance. Neither of the two arousalmediation theories can readily account for interactions between
extraversion and measured arousal, as these show effects of extraversion on performance with arousal held constant.
Method
Subjects
Subjects were 45 men and 71 women taking part in a study of the
acquisition of coding and touch-typing skills for an industrial (post
office) keyboard task. The men had a mean age of 31.2 (range 17-62),
the women a mean age of 33.8 (range 17-60).
EFFECTS OF EXTRAVERSION AND AROUSAL
AM
Fe.
631
PM
3.0--
II'IExtraverts
2,5 m
['~Ilntroverts
20--
Z
/
1.5--
/
/
1.0--
\
d
0.5-0.0
I
I
Low
High
I
I
I
Low
High
I
Energetic Arousal
Figure 1. Missed targets, in arbitrary normalized units, as a function of time of day, extraversion, and
energetic arousal. (Data from Matthews, in press. M = 1.42; SD = 1.25; N = 60.)
Design
Each subject completed all individual difference measures and cognitive tasks. Subjects were tested on the cognitive tasks at the same time
of day on each of the two days of testing. Seventeen men and 37 women
performed in the morning (10:00 AM), 28 men and 34 women in the
evening (8:00 PM). The basic design of the experiment was a 2 × 2 × 2
(Time of Day x Extraversion Level × Arousal Level) factorial design,
with median-split factors ofextraversion and arousal. In addition there
were within-subjects variables associated with three of the tasks used.
Measures of Individual Difference Variables
Measures of extraversion, impulsivity, and sociability were taken
from the Eysenck Personality Inventory (EPI; H. J. Eysenck & Eysenck,
1963). Energetic arousal, which Thayer (1978) saw as a measure ofreticulocortical arousal, was measured with the 48-item University of Wales
Institute of Science and Technology (UWlST) Mood Adjective Check
List (UMACL; Matthews, Jones, & Chamberlain, 1988). The UMACL is
an anglicization and extension of Thayer's (1967) Activation-Deactivation Adjective Check List (AI~ACL), whose Use of American English
adjectives such as peppy makes it unsuitable for use with British samples. Matthews, Jones, and Chamberlain (1988) showed that the UMACL
ener*detic arousal dimension was factorially distinct from two other dimensions of tense arousal and hedonic tone in a sample of 388 subjects.
The energetic arousal scale is composed of 5 of Thayer's energetic
arousal items and 10 new items. In the Matthews, Jones, and Chamberlain data the correlation between scales corresponding to the two item
sets is 0.80, which is comparable with the a of 0.90 for the whole scale.
Thus, the UMACE energetic arousal scale is equivalent to the Thayer
dimension. Matthews (1987) reported positive correlations between
UMACL arOusal scales and psychophysioiogleal measures of autonomic
arousal. Because of time constraints, we administered the UMACL on
the first day of testing only. Energetic arousal on the second day of testing
was estimated from a series of six UMACEadministrations, including the
first day of testing and five subsequent occasions during the keyboard
training program that followed this experiment and lasted for several
weeks. The a value for the six measures of energetic arousal was 0.84.
This high a suggests that where the testing environment remained approximately constant, individual differences in energetic arousal were
fairly stable. This stability of energetic arousal scores justifies the estimation of arousal on the second day of testing, although the lack of a
direct Day 2 measure is not ideal.
Task Apparatus and Stimuli
Tasks were controlled by a BBC Model B microcomputer, which was
also used for the recording and timing of responses. Tasks were semiautomated, such that instructions and practice sessions were administered by microcomputer.
Five-choice serial reaction. Responses for this task were made on
a special apparatus composed of a horizontal square surface with five
buttons forming the points of an equilateral pentagon with sides approximately 8 cm long. At the center of the pentagon was a sixth button.
At the rear of the apparatus was a surface inclined at an angle of approximately 45 °, with five lights arranged in a pentagonal configuration similar to that of the five buttons on the horizontal surface. When a light
was illuminated, the subject had to press the button corresponding in
position to that of the light. When the subject pressed any one of the
five buttons, the light was immediately switched off, and another (or
possibly the same one) was switched on. The subject used only one hand.
The task lasted for a fixed time of 15 min. Data were collected for each
half of the task, the main performance measures being the mean response latency of correct detections and the error rate (expressed as the
percentage of incorrect responses made).
Vigilance. This task was a microcomputer version of a task developed
by Nuechterlein, Parasuraman, and Jiang (1983). Single digits were presented on the visual display unit, degraded by a dot matrix film placed
on the screen. Stimulus display time was 50 ms, with a 1,000-ms onset
interval between successive stimuli. Subjects had to press the space bar
on the computer keyboard when they detected a zero. The test session
was made up of 540 stimuli ( 135 targets) and was preceded by a demonstration display and practice session of 30 stimuli. Nontarget digits ( 1-
632
G. MATTHEWS, D. JONES, AND G. CHAMBERLAIN
9) were presented with equal probability. Following presentation of a
target, from 1 to 5 nontargets were presented before the next target was
displayed. Neither targets nor individual nontargets were displayed
twice in succession. Data collected were hit rate and false-alarm rate for
each of three equal parts of the test session (180 stimuli, 45 targets). We
used standard tables to calculate the principal performance measures
d' and/3.
Digit span. This task required subjects to learn and recall 30 ninedigit lists. Each list was composed of the digits 1-9 with no digit repeated. A unique set of lists was generated for each subject. Stimuli were
double-height Teletext characters and were presented in the center of
the screen for 250 ms with a 1,000-ms interval between the onsets of
successive stimuli. After the final (ninth) digit in each list, a tone was
sounded, following which the subject had 15 s in which to recall the
digits. A second tone signaled that 3 s were left of the recall period.
Subjects made written responses on a prepared form that required the
subject to show the order in which the digits had appeared in the list.
Responses were scored as correct for each digit recalled in the correct
serial position. Thus, the primary performance measure was error probability per digit for each of the nine serial positions.
Posner et al. (1969) letter-match task. Subjects were presented with
pairs of letters in double-height Teletext and indicated, by pressing one
of two keys, whether the letters were the same or different. The stimuli
were pseudo-randomized pairings taken from the upper- and lowercase
of the letters A, B, D, E, and R. There were two kinds of similiarity of
letter pairs. Physical identity (PI) letter pairs were the same both in case
and letter name (e.g., A A); name identity (NI) pairs were the same only
in letter name (e.g., A a); and some pairs were different letters (e.g., B
a). There were a total of 320 stimuli: 80 PI pairs, 80 NI pairs, and 160
different letter pairs. Subjects were required to respond as soon as possible, with the stimulus remaining on screen until a response was made.
There was a 1,000-ms gap between stimuli during which a plus sign was
displayed in the same position as the letter pair. Data collected were
number of errors and mean response latency for each of the three types
of letter pair. The primary performance measures were latencies in the
PI and NI conditions and the difference in latencies NI minus PI, a
measure of speed of accessing letter name codes.
Six-letter search. Subjects were presented with a screen display in
double-beight uppercase, in the BBC Teletext display mode, composed
of an upper row of 6 letters, each separated by a space, and a lower row
of 20 randomly generated letters. The task was to indicate, by pressing
one of two dedicated keyboard keys, whether any of the 6 characters in
the upper row were missing from the 20 in the lower row. Subjects were
not told that not more than 1 could be missing in any display. They had
unlimited time to complete 5 practice and 25 test trials. Data collected
were number of errors and mean response latency.
Tapping tasks. We obtained five different tapping measures. Subjects
were required to tap keys on a microcomputer keyboard as fast as they
could for 10 s per measure. They were required to (a) use the index
finger of the nondominant hand to tap a single key, (b) use the little
finger of the dominant hand to tap a single key, (c) alternately use the
left index finger to tap one key and the right index finger to tap a second,
(d) use the index finger of the dominant hand to tap two keys separated
by nine keys, and (e) use the index finger oftbe dominant hand to tap
two keys separated by two keys. Mean latencies of tapping were measured. All five latencies were quite highly positively correlated (range of
rs, .32-.77). Hence, the mean latency of the five measures was used as
the primary measure of performance; this measure had an intermeasure reliability (a) of 0.81.
Procedure
On arrival on the first day of testing, the experimenter briefed subjects
and informed them that they were not expected to be skilled at key-
board or computer operation. Almost all subjects were tested in groups
of 2--4 individuals at the same time of day on two successive days. A
small number of subjects were tested alone due to nonappearance of
other subjects. The cognitive tests and the mood checklist were administered in an artificially lit room, with subjects separated by at least 2 m.
Subjects were seated so that other subjects were at least 90* from the
line of sight. On each day, subjects completed three of the six tests used.
To control for order effects, the tests were divided into two groups of
three (tapping; six-letter search; Posner et al., 1969, letter-match task;
and vigilance, memory, and serial reaction). Approximately half the
subjects performed the first group of tests on the first day and the second
group on the second day, whereas the remaining subjects performed the
two groups of tests in the opposite order. Within each group of three
tests, the order in which tests were performed on any one day was also
counterbalanced. The experimenter administered the first UMACL following the initial briefing, but prior to performing the cognitive tasks,
in the same room on the first day of testing. On the subsequent five
occasions during which subjects performed more complex keyboard
training tasks, the UMACL was also given prior to task performance in
the same room. Arousal measures were deliberately taken prior to performance because subjects' self-appraisals of performance might affect
arousal. Subjects completed the EPI prior to testing.
Results
Preliminary Analyses
M e a n s a n d s t a n d a r d deviations o f the variables used are
shown in Table 1. We used a log t r a n s f o r m o f the/3 values because they were positively skewed. T h e r e were n o significant
differences between the two times o f day in m e a n levels o f age,
gender, n u m b e r o f O' levels, or extraversion. Subjects were
d r a w n f r o m a range o f socioeconomic classes, p r e d o m i n a n t l y
(77%) Classes I (professional), II (intermediate), III N ( n o n m a n ual skilled), a n d III M ( m a n u a l skilled). T h e r e was n o significant
difference between the socioeconomic class frequency distribution at each t i m e o f day.
We c o m p u t e d Pearson p r o d u c t - m o m e n t correlations between extraversion a n d the Day 1 a n d estimated Day 2 energetic
arousal values. B o t h correlations were positive a n d significant:
0.21 for Day I arousal, 0.21 for Day 2 arousal ( N = 116, p <
.05). Correlations between extraversion a n d arousal at each
t i m e o f day were o f similar m a g n i t u d e to these values a n d did
n o t differ significantly. Correlations a m o n g impulsivity a n d sociability a n d arousal were likewise similar in magnitude. T h e
correlation between impulsivity a n d sociability here was 0.34.
Effects o f Time o f Day, Extraversion, and
Arousal on Performance
We tested effects o f i n d e p e n d e n t variables on p e r f o r m a n c e s
b y using analyses o f variance (ANOVAS).Between-subjects variables were t i m e o f day a n d median-split factors o f extraversion
a n d arousal (two levels each), with arousal split at the m e d i a n
for subject's m e a n arousal. For some tasks, we included withinsubjects variables for t i m e o n task (vigilance, serial reaction) or
serial position (digit span). All ANOVAS used a regression model
because o f u n e q u a l cell n u m b e r s . Day o f testing was n o t included in these analyses. W i t h i n each analysis, the arousal value
for each subject was set to the actual or estimated arousal value
for the day on w h i c h they were tested o n the task concerned. As
EFFECTS OF EXTRAVERSION AND AROUSAL
Table 1
Descriptive Statistics for Demographic, Independent,
and Dependent Variables
Variable
Demographic
Age
No. of O' levels
Independent
Extraversion
Impulsivity
Sociability
Energetic arousal (Day 1)
Energetic arousal (Day 2 estimate)
Dependent
Five-choice serial reaction: RT (ms)
Five-choice serial reaction: errors (%)
Vigilance: missed targets (%)
Vigilance: false alarms (%)
Vigilance: d'
Vigilance: log B
Digit span: error probability/item
Letter match: PI RT (ms)
Letter match: NI - PI RT (ms)
Letter match: PI errors (%)
Letter match: NI errors (%)
Six-letter search: RT (s)
Six-letter search: errors (%)
Tapping: latency (ms)
M
SD
32.8
4.9
10.6
3.6
I 1.4
4.7
6.2
47.2
45.0
4.2
2.0
2.5
8.3
7. l
693.1
1.7
8.2
6.4
3.63
-0.34
0.45
671.9
234.9
1.9
10.9
13.7
5.6
212.9
122.4
i.7
12.9
3.9
0.98
0.67
0.19
158.8
167.0
2.9
8.7
3.3
8.3
40.5
Note. RT = latency of correct responses; NI = name identity; PI = physical identity.
a result of this procedure, and of the counterbalancing of task
with respect to day of testing, the allocation of subjects to high
or low arousal groups varied slightly from task to task, although
each analysis had a total of 116 subjects. Cell means and Ns for
all these analyses are given in the Appendix.
Two of the tasks used appeared to be insensitive to the effects
of the individual difference variables. No significant effects on
tapping speed were found. The only effect on the six-letter
search task was a main effect of time of day on response speed,
F( 1, 108) = 6.5, p < .05; performance was faster in the evening.
Effects on the other tasks were as follows.
Five-choice serial reaction. We used mean response latency
and error rate as the dependent variables here, with task period
(two levels) as a within-subjects independent variable. We found
significant main effects on response latency of time of day, F ( I ,
108) = 6.8, p = .01; of arousal, F(1, 108) = 4.4, p < .05; and of
task period, F(1, 108) = 6.1, p < .05, together with significant
interactive effects of Time of Day × Extraversion × Arousal,
F(I, 108) = 7.5, p < .01, and of Arousal × Task Period, F(1,
103) = 7.9, p < .01). The relevant cell means are shown in Figure 2. Serial reaction was faster in the evening, in subjects low
in arousal, and in the second task period. However, the increase
in speed with time on task was shown only by high arousal
groups, who tended to be slower than low arousal groups in the
first task period. The triple interaction was associated with the
high arousal slowing introverts considerably but speeding extraverts slightly in the morning, whereas in the evening high
arousal tended to slow extraverts but speed introverts.
Only one effect of independent variables on error rate
633
reached significance, the main effect of task period, F ( I , 108) 9.7, p < .01. Mean error rate was less in the first period (1.55%)
than in the second (2.12%). The main effect of arousal on error
rate was close to significance, F ( I , 108) = 3.9, p = 0.052, with
higher arousal associated with fewer errors. Thus, both task period and arousal appeared primarily to affect the speed-accuracy tradeoff rather than efficiency of performance.
Vigilance. Dependent variables were d' and log ~. Task period was included in the analyses as a within-subjects variable
(three levels). We found no significant main or interactive
effects of independent variables on d', although there were
trends toward superior performance in high arousal subjects,
F ( I , 108) = 3.1, p < .10, and in the evening, F(I, 108) = 2.8,
p < .10. Significant effects on log ~ were as follows. Log O increased with task period, F(2, 216) = 15.7, p < .001), and means
for Periods 1-3 were -0.59, -0.29, and -0.13, respectively. Increases in criteria of this kind are commonly found with high
signal probability tasks (Davies & Parasuraman, 1982). The interactive effects on log/3 of Time of Day × Extraversion ×
Arousal X Task Period also reached significance,/7(2, 216) =
4.3, p < .05. In Period 2, high arousal lowered/~ in extraverts
and raised it in introverts in the morning but had the opposite
effect in the evening. Group differences in Periods 1 and 3 were
small.
Digit span task. Probability of recall failure per digit is shown
in Figure 3 as a function of the independent variables. The main
effect of time of day, F(1, 108) = 4.8, p < .05, reached significance, with worse recall in the morning. The Time of Day ×
Extraversion × Arousal interaction was also significant, F(I,
108) = 4.0, p < .05). In the morning, high arousal impaired
recall in introverts but slightly facilitated recall in extraverts. In
the evening, high arousal improved recall in introverts but had
little effect in extraverts.
Posner et al. (1969) letter match. We found no significant
effects on response time in the PI condition, but the interactive
effect of time ofday, extraversion, and arousal on the NI minus
PI index reached significance, F(I, 108) = 6. l, p < .05. In the
morning, high arousal increased speed of letter name access in
introverts but slowed access in extraverts. In the evening, both
groups were faster on this index when high in arousal, but the
effect of arousal was stronger in extraverts.
Effects of lmpulsivity and Sociability on Performance
As a supplementary analysis, we reran the ANOVAS,replacing
the extraversion variable with median-split factors ofimpulsivity and sociability. Our aim was to test whether effects of impulsivity were stronger than those of sociability, as previous research suggests (Revelle et al., 1980).
We found no significant main or interactive effects of either
impulsivity or sociability with the tapping or six-letter search
tasks. Significant effects were as follows (unless otherwise
stated, the direction of effects is the same as corresponding extraversion effects). Significant effects involving only sociability
were found with the vigilance task. An interactive effect on d'
of time of day, sociability, and arousal, F(1,108) = 6.4, p < .05,
was associated with high arousal raising d' in high sociables and
lowering d' in low sociables in the morning, while tending to
have the opposite effect in the evening. A significant interactive
634
G. MATTHEWS,
D. JONES, AND G. CHAMBERLAIN
AM
PM
850 -
800 -
750 -
g
700 -
--'°
Introvert/low arousal
Extravert/high arousal
13 . . . . . . . . . . .
650
Introvert/high arousal
IQ
-
Extravert/low arousal
600
I
1
I
2
I
I
1
I
I
2
Period
Figure2. Speed of five-choice serial reaction (ms) as a function of
time of day, extraversion, and energetic arousal.
effect of time of day and sociability on log/3, F(I, 108) = 4.3,
p < .05, was associated with a positive association between sociability and criterion in the morning, with the opposite association in the evening. Effects of impulsivity and sociability on
mean response latency in the five-choice serial reaction task
were similiar to those of extraversion. Both the Time of Day X
Sociability × Arousal interaction, F ( l , 108) = 4.8, p < .05, and
the Time of Day × Impulsivity x Arousal interaction, F(1,
108) = 8.7, p < .01, reached significance. However, although
sociability did not affect error rate, there was a significant main
effect of impulsivity on error rate, F(I, 108) = 6.7, p < .05,
associated with higher error rates in high impulsives. This main
0.60 -
0.50Introvert/low arousal
Introvert/high arousal
0.40 -
Extravert/low arousal
Extravert/high arousal
h
0.30-
0.20
I
I
AM
PM
Figure 3. Error probability per item in digit span task as a function of
time of day, extraversion, and energetic a r o u s a l .
EFFECTS OF EXTRAVERSION AND AROUSAL
effect was modified by a significant Impulsivity × Arousal interaction, F ( l , 108) = 4.3, p < .05. Adverse effects of impulsivity
on accuracy were most marked under high arousal. We found
no significant effect of impulsivity on errors in the digit span
task. However, the interactive effect of sociability and time of
day on this task reached significance, F ( l , 108) = 4.2, p < .05):
High sociables had relatively better recall in the evening. The
triple interactive effect of time of day, personality, and arousal
on the letter-match NI minus PI index reached significance for
sociability, F(I, 108) = 3.9, p = .05, and was close to significance for impulsivity, F(1, 108) -- 3. l, p < . 10. To summarize,
these analyses suggest that, if anything, interactions involving
time of day, sociability, and arousal were more prevalent than
those involving time of day, impulsivity, and arousal.
Tests for Curvilinear Regressions of Arousal and
Extraversion on Performance
As a further supplementary analysis, we tested curvilinear
regressions of arousal on performance of the kind predicted by
the Yerkes-Dodson law. For each dependent variable, we performed a multiple regression, including arousal (linear term)
and arousal squared (quadratic term) as predictors in the equation. We converted both linear and quadratic terms to z scores
to reduce the correlation between them. If arousal were related
to performance by an inverted IJ, the quadratic term should
make a significant individual contribution to the equation, and
the B weight of the term should be negative in sign. For tasks in
which task period was treated as a within-subjects variable, we
computed an average value for the dependent measures concerned. These measures were d' and log B from the vigilance
task and response latency and error rate from the five-choice
serial reaction task. None of the multiple regressions were significant overall (and none of the F values for the unique contributions of the quadratic terms to the regressions reached significance). Thus, the data fail to provide any evidence for inverted-LI-shaped regressions of performance on self-report
arousal within this experiment.
We also performed similar tests for curvilinear effects of extraversion on arousal. Only one quadratic term reached significance: The square of extraversion was negatively related to response time in the PI condition of the Posner task, F ( I , 112) =
5.2, p < .05.
Tests for Artifacts of Gender and Order of Presentation
The final set of analyses tested for artifactual effects of gender
and order of presentation on the results presented. There are
two ways in which gender might influence the results. First, gender might modify the effects ofextraversion and arousal on performance. We did not perform tests of such effects for the following reasons: There is no obvious theoretical rationale for
predicting such interactive effects of gender and other variables
on performance, previous research in this area has used both
single- and mixed-gender groups, and there were insufficient
subjects to test four-way interactions between gender and the
three between-subjects variables. Second and more important,
if gender is correlated with both performance and with main
effect or interaction terms in the ANOVA models of the main
635
analyses, then artifacts may result. To test this possibility, we
tested the effects of gender and time of day on each dependent
variable within a fully factorial 2 × 2 ANOVA. We included time
of day because of the somewhat unequal proportions of the two
genders within the morning and evening groups. There were no
significant gender differences in extraversion or average energetic arousal. We found only one significant gender difference:
men were faster at tapping, F ( I , 112) = 3.98, p < .05. As there
were no gender differences in performance on the tasks sensitive
to extraversion and arousal, these individual difference effects
cannot be an artifact of gender.
Similarly, order of testing might influence results if order
both affected performance and was correlated with the terms
in the ANOVA model. We performed further ANOVAS for each
dependent variable, with day of testing (two levels) and position
in the test sequence on the day of testing (three levels) as between-subjects variables. The only task that was sensitive to
such effects was the six-letter search. On the six-letter search,
more errors were made in the third position in the sequence
than in the earlier positions, F(2, 107) = 3.33, p < .05, but response latencies were somewhat faster in the third position, F(2,
107) = 2.98, p = .055. We found no significant effects of day of
testing or Day × Position interactions. Hence, the significant
interactive effects of time of day, extraversion, and arousal that
we found cannot have been an artifact of order effects on performance.
Discussion
Tests of Predictions From Arousal-Mediation
Models of Personality
In the first section of the discussion we consider the success
of the predictions made if the arousal measure is assumed to be
valid. We discuss possible difficulties with the validity of arousal
in the next section.
1. The relation between extraversion and arousal. In this
study, extraversion was positively correlated with arousal irrespective of time of day, although the magnitude of the relation
was small. This relation was not predicted by either the H. J.
Eysenck (1967) or the Humphreys and Revelle (1984) theory;
however, it is consistent with Matthcws' (1987) review of relevant studies. A positive correlation between extraversion and
arousal is compatible with the H. J. Eysenck (1967, 1981) theory if it is accepted that experimental conditions were sufficiently stimulating to lower arousal in introverts by transmarginal inhibition. However, under typical experimental conditions, if extraverts are normally more aroused than introverts,
it is difficult to see why extraverts' performance in the morning
tends to benefit from external stressors.
2. The relation between arousal and performance. In general, task performance was fairly insensitive to main effects of
arousal. Only with five-choice serial reaction was there an effect
of arousal independent of extraversion, and this appeared to be
an effect on the speed-accuracy tradeoff rather than on efficiency. High arousal tended to make performance slower but
more accurate. Given that arousal-mediation models of personality see arousal as affecting performance directly, but see extraversion as affecting performance only indirectly through its cor-
636
G. MATTHEWS, D. JONES, AND G. CHAMBERLAIN
relation with arousal, the weakness of main effects of arousal in
this study is surprising from the perspective of these models.
The analysis of quadratic trends in the regression of performance on arousal suggests that, within the range of applicable
energetic arousal values, the Yerkes-Dodson law is a poor predictor of relations between self-report arousal and performance.
3. Interactive effects of extraversion, arousal, and performance. We found the predicted interactive effects of time of'
day, extraversion, and arousal with two tasks: speed of serial
reaction and digit span. Both triple interactions were of the
form expected on the basis of previous research (e.g., Matthews,
1985). High arousal tended to facilitate extraverts' performance
in the morning but introverts' performance in the evening. Although these interactions are conceptually similar to those reported by Revelle et al. (1980) with impulsivity and caffeine,
they appear to be incompatible with any model positing a linear
causal chain such that personality affects arousal, which in turn
affects performance, because of the interaction between extraversion and directly measured arousal.
Interactive effects of the three between-subjects independent
variables on the NI minus PI measure from the letter-match
tasks were in the opposite direction to the usual finding. Extraverts showed faster speed of access to letter name codes when
low in arousal in the morning but high in arousal in the evening.
This finding is hard to explain from any existing personality
theory.
A further difficulty for the Humphreys and Revelle (1984)
theory is the failure to replicate the finding that Extraversion ×
Arousal interactions are mediated by impulsivity. One contributory factor may be the moderate reliability of the EPI Impulsivity scale. Matthews, Davies, and Lees (1988) report a values
for EPI extraversion, sociability, and impulsivity of.77, .69, and
.63, respectively (N = 100 undergraduates). The task-specific
effects ofimpulsivity on serial reaction, showing that impulsive
subjects tend to be error-prone (particularly when high in
arousal), are consistent with choice reaction-time data reported
by Edman, Schalling, and Levander (1983).
Accommodation of the Data Within Arousal-Mediation
Models
Arousal-mediation models can only account for the interaction between extraversion and a direct measure of arousal if
self-report energetic arousal is not actually a valid measure of
cortical arousal. One problem is that the validity of the arousal
measure may have been affected by environmental influences,
particularly the testing of subjects in small groups and the overall level of stress. Group testing may affect arousal level; however, in a similar experiment on individually tested subjects
(Matthews, in press), mean pretest energetic arousal was 44.6,
only slightly less than mean levels in the present experiment.
The overall level of stress or arousal may affect extraversionarousal correlations through transmarginal inhibition although
in this study it is unlikely that the experiment was more than
mildly stressful. The use of a traitlike measure to estimate the
Day 2 arousal state may have lowered the reliability of the Day
2 measure. However, because the UMACL was administered in
a similar context on all six occasions, it is unlikely that the estimation procedure was a source of systematic error.
A statistical difficulty is the possibility of restriction of range
of arousal scores. Subjects of more extreme arousal levels than
those sampled here might well show the performance decrements predicted by the Yerkes-Dodson law. However, because
we actually found interactive effects of arousal and extraversion,
any restriction of range can only be moderate. Data reviewed by
Thayer (1978) show that commonly used arousal manipulations
have effects of moderate magnitude on self-report arousal: typically about 1-2 standard deviations with overlap between the
arousal distributions in high- and low-arousal groups resulting
from individual differences in arousal. Here, the mean average
arousal levels of high- and low-arousal groups (51.4 and 39.6)
differed by 1.65 standard deviation. Thus, analysis of groups
median-split on self-report arousal, as in this study, will be of
similar power to analysis of high- and low-manipulated-arousal
groups.
Other evidence suggests that adjective checklist measures are
likely to be valid under performance testing conditions. Thayer
(1967) and Matthews (1987) both found significant positive correlations between self-report and psychophysiological arousal
measures in testing contexts. Matthews, Davies, and Lees
(1988) computed the test-retest correlations of energetic
arousal measures taken before and after attentional task performance. In three experiments, these correlations were .62, .68,
and .76, showing reasonably stable individual differences in
arousal levels. In summary, the nature of the experimental environment, possible restriction of range, and changes in arousal
over time are unlikely to have had serious adverse effects on
validity.
Perhaps a more serious problem is lack of evidence allowing
self-report arousal measures to be linked to specific physiological arousal systems, such as H. J. Eysenck's (1967) reticulocortical loop. It is possible that manipulations of arousal, such as
the caffeine manipulation used by Revelle et al. (1980), are in
fact affecting a different arousal system from that measured by
self-report energetic arousal. Thayer's (1978) validation evidence could perhaps be explained if self-report arousal is a direct measure of autonomic arousal, which in turn tends to generate cortical arousal, as in the H. J. Eysenck (1967) theory. If
total cortical arousal was the sum of autonomically generated
cortical arousal plus cortical arousal associated with personality, then the data would become more compatible with the
HumPhreys and Revelle (1984) model. High autonomic
arousal, as measured by self-report, would lead to overarousal
in introverts in the morning but would lead to overarousal in
extraverts in the evening. Correlations between extraversion
and autonomic arousal, and effects of autonomic arousal on
performance, might then be considered epiphenomenal to such
effects of extraversion.
There remains the problem of the role of task factors in controlling interactive effects ofextraversion and arousal on performance. The assumption made by Humphreys and ReveUe
(1984) is that interactions of the kind shown in Figures 1-3 reflect an underlying inverted-U regression between arousal and
performance. However, their model generates such regressions
only for complex tasks: those drawing on both STM and SIT
resources. The complex task here--six-letter search--was in-
EFFECTS OF EXTRAVERSION AND AROUSAL
sensitive to effects ofextraversion and arousal. As pure memory
and pure sustained attention tasks, respectively, digit span and
serial reaction tasks should be related to arousal by monotonic
regressions, which cannot generate the interactions shown in
Figures 1-3. Thus, to account for the interaction data using the
Humphreys and Revelle model, the assumption would have to
be that serial reaction and digit span are complex tasks. Conversely, the six-letter search task, with a higher error rate and
mean response time than serial reaction, could not be limited
by the two processing resources. Evidence for a possible attentional component of the digit span task is discussed later, but it
is difficult to argue for a STM component in serial reaction.
To summarize, neither of the arousal-mediation theories
were successful in predicting relations between the three principal independent variables and performance. The data can only
be accommodated within the Humphreys and Revelle (1984)
model by making uncorroborated assumptions about (a) the
relation among extraversion, self-report arousal, and cortical
arousal, and (b) the resources required by the tasks used.
Extraversion, Arousal, and Stimulus Identification
Predictions from Matthews's work (1985, in press) were partially successful in that interactive effects of extraversion and
self-report arousal on two of the tasks were consistent with previous research. However, if these variables affect stimulus identification, several of our results are surprising: the lack of a significant effect of time of day, extraversion, and arousal on vigilance task, and the significant interactive effects of these
variables on the two memory tasks of digit recall and letter
matching. The lack of significant effects on d' in the vigilance
task could be a chance effect as the triple interaction showed a
trend in the predicted direction that reached significance with
the sociability subfactor. Studies of noise (e.g., Davies &
Hockey, 1966) and of caffeine (Keister & McLaughlin, 1972)
have shown characteristic interactive effects of extraversion and
manipulated arousal on vigilance. Also, the analysis of/~ effects
suggested a complex effect of extraversion and other independent variables on task strategy, which may have masked effects
on sensitivity. Dempster (1981) concluded that a major source
of individual differences in digit span tasks is speed of item identification. The personality effects on digit recall perhaps operated on this attentional component of digit span rather than on
short-term storage per se. The significant interactive effect of
time of day, extraversion, and arousal on the NI - PI measure
from the letter-match task was in the opposite direction to the
typical interaction between these variables. Possibly, the conscious application of a verbal label to a letter takes place relatively late in stimulus analysis, after lower level processes have
analyzed simple physical and semantic features (see Broadbent,
1982).
Thus, Matthews's (in press) characterization of extraversion
and arousal as interactively operating on lower-level stimulus
encoding processes captures two aspects of such personality
effects that arousal-mediation theories do not. First, extraversion and arousal effects appear to be genuinely interactive and
irreducible to an inverted-O regression. Second, such interactions are most prevalent with tasks whose performance depends
on simple stimulus encoding. Two further issues must be con-
637
sidered. First, can the stimulus encoding hypothesis account for
the wide range of tasks subject to interactive effects of extraversion and arousal? Second, can the mechanism of encoding
affected be specified more precisely? These questions are addressed in the remainder of this article.
Other personality research supports the stimulus encoding
hypothesis. Examples of pure detection tasks showing the characteristic Extraversion X Arousal interaction are Blake's ( 1971)
studies of letter cancellation and vigilance studies reviewed by
Eysenck (1982). Anderson and Revelle's (1982) study ofimpulsivity, caffeine, and proofreading is another possible example.
In contrast, two studies of attentional tasks requiring the integration of short-term storage and attention (Anderson & Revelle, 1983; M. W. Eysenck & Eysenck, 1979) failed to find significant Personality X Arousal interactions. The encoding hypothesis can be extended to tasks not generally classified as
attentional. M. W. Eysenck's (1981) studies of retrieval, which
show a typical Extraversion X Arousal interaction, require generation of words in response to a cue word. Cue encoding may
well be an important determinant of subsequent retrieval. Verbal ability tests have provided some of the most consistent evidence for the arousal dependency of effects of extraversion
(Matthews, 1985) and impulsivity (ReveUe et at., 1980) on performance. Such effects could reflect a variety of mechanisms.
The present hypothesis is consistent with the association between intelligence and speed of encoding simple stimuli (Vernon, 1986). Effects of extraversion on intelligence may result
from subjects' initial encoding of stimulus material rather than
from the subsequent processing necessary for solution of the
item.
Finally, as encoding is not a unitary process, what specific
encoding mechanism is interactively affected by extraversion
and arousal? To encompass the semantic memory and ability
test data, the mechanism must be sufficiently late in processing
to affect the development of a semantic code for the stimulus
material. That is, the mechanism is attentional rather than perceptual in nature. However, the mechanism must not be so late
that effects would be expected on a complex search requiring
the controlled coordination of attention and working memory,
such as a six-letter search, or on the (unpracticed) Posner et at.
(1969) letter-match task. Hence, the most likely locus of personality effects is the preselective encoding of simple semantic attributes of the stimulus. Semantic priming phenomena demonstrate that such encoding can take place prior tO later stages
of analysis required for processes of selection on the basis of
complex semantic attributes, conscious identification, and controlled processing (Dark, Johnston, Myles-Worsley, & Farah,
1985). This incomplete stimulus analysis may be sufficient to
allow the fairly simple discriminations between targets and nontargets typical of the attentional tasks sensitive to interactive
effects of time of day, extraversion, and arousal.
References
Anderson, K. J., & Revelle, W. (1982). Impulsivity, caffeine, and proofreading: A test of the Easterbrook hypothesis. Journalof Experimental Psychology:Human Perceptionand Performance, 8, 614-624.
Anderson, K. J., & Revelle, W. ( 1983). The interactive effectsof caffeine,
impulsivity and task demands on a visual search task. Personalityand
Individual Differences. 4, 127-134.
638
G. MATTHEWS, D. JONES, AND G. CHAMBERLAIN
Baddeley, A. (1986). Working memory Oxford, England: Oxford University Press.
Blake, M. J. E (1971). Temperament and time of day. In W. E Colquhoun (Ed.), Biological rhythms and human behaviour (pp. 109148) London: Academic Press.
Broadbent, D. E. (1971 ). Decision and stress. London: Academic Press.
Broadbent, D. E. (1982). Task combination and the selective intake of
information. Acta Psychologica, 50, 253-290.
Clubley, M., Bye, C. E., Henson, T. A., Peck, A. W., & Riddington, C. J.
(1979). Effects of caffeine and cyclizine alone and in combination
on human performance, subjective effects and EEG activity. British
Journal of Clinical Pharmacology, 7, 157-163.
Corcoran~,D. W. J. (1965). Personality and the inverted-U relation. British Journal of Psychology, 56, 267-273.
Dark, V. J., Johnston, W. A., Myles-Worsley, M., & Farah, M. J. (1985).
Levels of selection and capacity limits. Journal of Experimental Psychology: General, 114, 472-497.
Davies, D. R., & Hockey, G. R. J. (1966). The effects of noise and doubling the signal frequency on individual differences in visual vigilance
performance. British Journal of Psychology, 5 7, 381-389.
Davies, D. R., & Parasuraman, R. (1982). The psychology of vigilance.
London: Academic Press.
Dempster, F. N. ( 1981). Memory span: Sources of individual and developmental differences. Psychological Bulletin, 89, 63-100.
Duffy, E. (1957). The psychological significance of the concept of
'arousal' or 'activation'. Psychological Review, 64, 265-275.
Edman, G., Schalling, D., & Levander, S. E. (1983). Impulsivity and
speed and errors in a reaction time task: A contribution to the construct validity of the concept of impulsivity. Acta Psychologica, 33,
1-8.
Eysenck, H. J. (1967). The biological basis of personality Springfield,
IL: Thomas.
Eysenck, H. J. (1981). General features of the model. In H. J. Eysenck
(Ed.), A model for personality (pp. 1-37). Berlin, West Germany:
Springer-Verlag.
Eysenck, H. J., & Eysenck, S. B. G. (1963). The Eysenck Personality
Inventory London: University of London Press.
Eysenck, M. W. (1977). Human memory: Theory, research and individual differences. Oxford, England: Pergamon.
Eysenck, M. W. (1981). Learning, memory and personality. In H. J.
Eysenck (Ed.), A model for personality (pp. 169-209). Berlin, West
Germany: Springer-Verlag.
Eysenck, M. W. (1982). Attention and arousal. Cognition and performance. Berlin, West Germany: Springer-Verlag.
Eysenck, M. W., & Eysenck, M. C. (1979). Memory scanning, introversion-extraversion, and levels of processing. Journal of Research in
Personality, 13, 305-315.
Humphreys, M. S., & Revelle, W. (1984). Personality, motivation and
performance: A theory of the relationship between individual differ-
ences and information processing. Psychological Review, 91, 153184.
Jones, D. M., & Coyle, K. (1983). Time of day and performance: A
levels of control approach. Paper presented to the British Psychological Society Annual Conference, York, England.
Keist~ M. E., & McLaughlin, R. J. (1972). Vigilance performance related to extraversion-introversion and caffeine. Journal of Experi-
mental Research in Personality, 6, 5-1 I.
Matthews, G. (1985). The effects of extraversion and arousal on intelligence test performance. British Journal of Psychology, 76, 479-493.
Matthews, G. (1987). Personality and multidimensional arousal: A
study of two dimensions of extraversion. Personality and Individual
Differences, 8, 9-16.
Matthews, G. (in press). Extraversion and levels of control of sustained
attention. Acta Psychologica.
Matthews, G., & Chappelow, J. E. (1986). Effects of extraversion and
arousal on selective attention. Paper presented to the British Psychological Society Cognitive Section Third Annual Conference, Oxford,
England.
Matthews, G., Davies, D. R., & Lees, J. L. (1988). Arousal, extraversion
and individual differences in resource availability. Manuscript submitted for publication.
Matthews, G., Jones, D. M., & Chamberlain, A. G. (1988). Refining
the measurement of mood: The UWIST Mood Adjective Checklist.
Manuscript submitted for publication.
Nuechterlein, K., Parasuraman, R., & Jiang, Q. (1983). Visual sustained attention: Image degradation produces rapid sensitivity decrement over time. Science, 220, 327-329.
Paap, K. R., & Ogden, W. G. (1981). Letter encoding is an obligatory
but capacity-demanding operation. Journal of Experimental Psychology."Human Perception and Performance, 7, 5 ! 8-528.
Parasuraman, R. (1985). Sustained attention: A multifactorial approach. In M. I. Posner & O. S. M. Marin (Eds.), Attention andperformance(Vol. I l, pp. 493-512). Hillsdale, NJ: Erlbaum.
Posner, M. I., Boies, S. J., Eichelman, W. H., & Taylor, R. L. (1969).
Retention of visual and name codes of single letters. Journal of Exper-
imental Psychology Monograph, 79, 1-16.
Revelle, W., Humphreys, M. S., Simon, L., & Giililand, K. (1980). The
interactive effect of personality, time of day, and caffeine: A test of the
arousal model. Journal of Experimental Psychology: General, 109, I 31.
Thayer, R. E. (1967). Measurement of activation through self-report.
Psychological Reports, 20, 663-678.
Thayer, R. E. (1978). Toward a psychological theory of multidimensional activation (arousal). Motivation and Emotion, 2, i-33.
Thayer, R. E. (1986). Activation-Deactivation Adjective Checklist: Current overview and structural analysis. Psychological Reports, 58,
607-614.
Vernon, E A. (1986). Inspection time: Does it measure intelligence?
Personality and Individual Differences, 7, 715-720.
639
EFFECTS OF EXTRAVERSION AND AROUSAL
Appendix
Table A- 1
Table A-2
Five-Choice Serial Reaction Performance as a Function
of Time of Day, Extraversion, Energetic
Arousal, and Task Period
Vigilance Task Performance as a Function of Time of Day,
Extraversion, Energetic Arousal, and Task Period
AM
Group
N
Extraverts
High arousal
Period l
Period 2
Low arousal
Period I
Period 2
Introverts
High arousal
Period I
Period 2
Low arousal
Period 1
Period 2
RT
AM
PM
Error
N
RT
Group
Error
16
16
736
707
1.34
1.82
17
17
703
688
0.91
1.43
8
8
740
729
2.42
2.52
15
15
616
618
2.34
3.06
11
11
818
780
1.14
1.99
15
15
686
668
1.13
1.62
19
19
661
662
1.02
1.29
15
15
685
729
1.59
2.11
Note. RT = mean latency of correct responses (ms). Error = mean error
Extraverts
High arousal
Period 1
Period 2
Period 3
Low arousal
Period l
Period 2
Period 3
Introverts
High arousal
Period 1
Period 2
Period 3
Low arousal
Period 1
Period 2
Period 3
PM
N
d'
log fl
N
d'
log fl
16
16
16
3.60
3.74
3.76
-0.43
-0.33
0.05
16
16
16
3.98
3.54
3.56
-0.76
-0.34
-0.37
8
8
8
3.29
2.87
3.14
-0.60
0.13
-0.17
16
16
16
3.56
3.59
3.38
-0.69
-0.47
-0.21
11
11
11
3.72
3.36
3.31
-0.67
-0.20
-0.04
14
14
14
4.21
4.42
4.10
-0.44
-0.38
-0.06
19
19
19
3.63
3.55
3.45
-0.66
-0.48
-0.18
16
16
16
3.68
3.68
3.52
-0.43
0.03
-0.02
rate (%).
Table A-4
Posner Letter Match Task Performance as a Function of Time
of Day, Extraversion, Energetic Arousal, and Task Period
Table A-3
Digit Span Task Performance as a Function of Time of Day,
Extraversion, Energetic Arousal, and Task Period
AM
Group
N
Extraverts
High arousal
Lowarousal
Introverts
High arousal
Lowarousal
Group
PM
Error
probability
per item
Error
probability
per item
N
AM
17
7
.504
.536
16
16
.378
.385
12
18
.577
.375
15
15
.440
.488
Extraverts
High arousal
Low arousal
Introverts
High arousal
Low arousal
N
PM
PI
NI - PI
N
PI
NI - PI
14
l0
673
677
329
194
23
9
655
682
206
328
14
16
718
691
213
279
14
16
644
652
179
191
Note. PI = correct mean response latency (RT) in physical identity condition (ms); NI - PI = difference between correct RTs in name and
physical identity conditions (ms).
Table A-6
Tapping Performance as a Function of Time of Day,
Extraversion, Energetic Arousal, and Task Period
Table A-5
Six-Letter Search Task Performance as a Function of Time of
Day, Extraversion, Energetic Arousal, and Task Period
AM
Group
Extraverts
High arousal
Lowarousai
Introverts
High arousal
Lowarousal
N
RT
PM
Mean error
rate (%)
N
RT
Mean error
rate (%)
15
9
14.4
14.3
9.08
6.24
22
lO
13.4
12.5
3.44
5.20
14
16
15.1
14.1
4.84
5.24
14
16
12.8
12.9
6.56
5.52
Note. RT = mean latency of correct responses (s).
AM
Group
Extraverts
High arousal
Low arousal
Introverts
High arousal
Low arousal
N
PM
Latency
N
Latency
15
9
220
220
23
9
205
213
14
16
211
210
14
16
205
226
Note. Latency = mean latency of tapping averaged across five tasks.
Received August 5, 1987
Revision received May 17, 1988
Accepted S e p t e m b e r 8, 1988 •