1) Levels of analysis and the psychological spectrum. Psychological phenomena
occur across at least 12 orders of temporal magnitude. Cognitive and
motivational theories at each frequency make use of directional and energetic
constructs. Outcome measures may be organized in terms of their temporal
resolution as well as their physiological emphasis. (Adapted from Revelle,
1989).
For psychologists concerned with linking cognition to action, it is essential to consider how motivational variables affect the competence-performance relationship. Ever since Blodgett's (1929) demonstration that well fed rats will learn mazes but that only hungry rats will show their knowledge by running rapidly through the maze, psychologists have been aware that competence is a necessary but not sufficient determinate of performance. An even more important study was Yerkes and Dodson's demonstration (1908) that motivational intensity (induced by foot shock) has a non-monotonic affect upon rates of learning a discrimination task and that task difficulty interacts with intensity.
Unfortunately many cognitive psychologists pay only lip service to the competence-performance distinction and will report that their subjects are well motivated and thus it is not necessary to worry about motivation. For such researchers, motivation is a nuisance variable that can be ignored by increasing sample size. The possibility that individual differences in personality might interact with situational manipulations in ways that can completely obscure important relationships is so foreign as to not even be considered.
An exception to this rule is those who have worked with or been inspired by Donald Broadbent. The best work on the effect on cognitive performance of non-cognitive manipulations such as noise, time of day, distraction, and incentives has been done by those who have followed the traditions established at the Applied Psychology Unit in Cambridge and continued at Oxford. Discussing motivational and stress effects before such a group is equivalent to bringing coals to Newcastle.
The emphasis of much of the work at the APU has been how stressors combine to affect performance[3]. Within this tradition, there has been great concern with the similarity and differences between the effects of different stressors. So, for example, while the effect of sleep deprivation is to hinder certain tasks, and noise to hinder other tasks, the combination of the two stressors can be shown to facilitate performance. An explanation that subsumes both effects is then proposed, tested and accepted or rejected (Broadbent, 1971).
This logic can equally well be applied to the combination of stressors with dimensions of individual differences. By appropriate analysis of the similarities and differences of effects due to experimental manipulations and individual differences it is possible to evaluate the construct validity of both. Certain individual differences seem to parallel certain stress manipulations while other stressors seems to affect different individuals in different ways. Both patterns of results are of theoretical importance: Parallel effects of personality and situational manipulation allow individual differences to be used to extend the effective range of experimental manipulations; different patterns for different people produce better theory by delineating the boundaries of effects of theoretical constructs.
Parallel effects of individual differences and situational stressors can suggest that both reflect differences on the same latent construct. By appropriate combinations of subject differences and of experimental manipulations, it is then possible to achieve a much greater effective range on the underlying latent construct than would be possible by manipulation or subject selection alone..[4]
There are at least three possible reactions to the observation that what improves the performance of one individual hinders the performance of another: 1) ignore that particular manipulation because it does not have consistent effects; 2) run more subjects in the hope that error terms will be reduced; or 3) ask what are the special characteristics of the different kinds of subjects. It is this third approach that is most useful. Understanding how manipulations differ across people leads to better theories of those manipulations as well as better theories of individual differences in personality. [5]
Individual differences in motivation and performance
Two dimensions of personality discussed by Broadbent (1958) as important sources of variation in performance were introversion-extraversion and stability-neuroticism. Extraversion was associated with decrements in performance over time and neuroticism was associated with greater decrements following stress. Although it is tempting to propose a single model to account for these effects, what has become clear is that the effects of personality upon performance require multiple levels of explanation. The broad dimensions of personality that are consistently identified from investigator to investigator and shown to be important in different cultures and different times affect behavior in many different ways.Before reviewing specific effects of personality, it is necessary to consider what are the appropriate dimensions to discuss. Personality researchers can be grouped into those who who study the effect of a single dimension versus those who develop taxonomic models of multiple dimensions. The first approach results in alphabetic organizations of personality traits (ranging from Type A behavior, through Machievelianism to Sensation Seeking) with numerous studies of convergent validity but few studies of discriminant validity. Within the second, multivariate-taxonomic tradition are those most concerned with description and those interested in causal (usually biological) theories. The descriptive taxonomists have agreed that a set of five dimensions can be identified consistently across methods. These "big five" dimensions of self report and peer description have been labeled Surgency, Agreeableness, Emotional Stability, Conscientiousness, and Culture, (Digman, 1990, Fiske, 1949, Goldberg, 1982; McCrae and Costa, 1987; Norman, 1963; and Wiggins, 1979). For the more biologically minded, the theories of Hans Eysenck (1952, 1967, 1981, 1991), Jeffrey Gray (1972, 1981, 1982, 1987), or Jan Strelau (1983, 1985) are appealing descriptions of three fundamental dimensions (Eysenck's Extraversion, Neuroticism, and Psychoticism; Gray's Impulsivity, Anxiety, and Aggression; Strelau's strength of the excitory and inhibitory processes, and balance between these process) that fit within the five factor model [6]. Whether one prefers the three dimensional biological models or the five dimensional semantic descriptions, it is clear that all of these dimensions have substantial genetic loadings and that they are moderately consistent from childhood throughout the life span.[7]
Perhaps because of a greater concern for causal theory among the biologically oriented taxonomists, there has been more research relating introversion-extraversion and stability-neuroticism to performance than there has been for the other dimensions of the "big five". Both of these dimensions may be associated with individual differences in motivational state. Although staying within the two-space defined by Introversion-Extraversion and Neuroticism-Stability, some of the more recent work has examined impulsivity, a component of I/E and anxiety, a component of neuroticism.
Motivational states: Affective valence and intensity
A common assumption when studying human performance is that subjects are alert and optimally motivated. It is also assumed that the experimenter's task at hand is by far the most important thing the subject has to do at that time. Thus, although individual differences in cognitive ability are assumed to exist, differences in motivation are ignored. For compliant college students participating in one of only a few psychology experiments, this assumption might well be true. It is probably less true for psychiatric patients, oil platform workers at the end of their shift, or deep sea divers under several hundred feet of water. Indeed, for almost any subject population of interest it is difficult to believe that the specific experimental task used has an equally powerful motivation effect upon all subjects. In fact, it is possible, even with college students, to show that variations in motivational state are important sources of between subject variation in performance.Motivational states can be categorized in several different ways. Conventionally, it has been useful to distinguish between the affective direction and the energetic intensity of motivation (Humphreys and Revelle, 1984). More recent work on affective states, however, has suggested that direction may subdivided into positive and negative components (Watson and Tellegen, 1985) and that intensity should be considered in terms of energetic and tense arousal (Thayer, 1989). How these four constructs interrelate is far from clear. Table 1 presents sample adjectives associated with each construct.
Table 1: Adjectives associated with the measurement of affect and arousal
Thayer's dimensions of arousal Watson and Tellegen dimensions of affect
Energetic Arousal Tense Arousal Positive Affect Negative Affect
energetic fearful alert nervous
full-of-pep jittery active jittery
active tense excited afraid
wakeful clutched-up enthusiastic scared
lively intense attentive guilty
vigorous (not) quiescent interested hostile
wide-awake (not) quiet inspired distressed
(not) sleepy (not) placid determined ashamed
(not) drowsy (not) still proud upset
(not) tired (not) at-rest strong irritable
(not) calm
An alternative four dimensional model of affect and arousal
High Energetic Low Energy/Tension High Depression High Tension
alert drowsy unhappy nervous
full-of-pep dull gloomy jittery
active placid blue afraid
wakeful quiet sad tense
lively serene depressed scared
aroused sleepy angry guilty
excited calm irritable surprised
Affective States
Thayer (1967, 1978, 1989) has discussed four uni-polar dimensions that he groups into two higher order constructs of energetic and tense arousal. He associates energetic arousal with approach behavior and tense arousal with avoidance behavior. Energetic arousal is increased by mild exercise and varies diurnally. Thayer (1989) adopts Gray's hypothesis that approach motivation reflects a sensitivity to cues for reward and that avoidance behavior reflects a sensitivity to cues for punishment. (See also Fowles, 1980).Matthews, Jones & Chamberlain (1989) report three mood dimensions that are sensitive to external stressors: energetic arousal, tense arousal, and hedonic tone (positive versus negative). They show that energetic arousal is decreased by the administration of Chlorpromazine, Diazepam or sleep deprivation. Tense arousal is increased by pain, or watching TV violence, but is reduced by muscle relaxation.
Watson and Tellegen (1985) have shown that positive and negative affect are independent of each other and can be used in combination to describe many psychopathological conditions. Clark and Watson (1991) recently proposed that differences in positive affect and somatic arousal account for the important distinction between two affective conditions represented by high negative affect, anxiety and depression. They suggest that while depression and anxiety share high negative affect, anxiety also reflects high somatic arousal, and depression represents lack of positive affect.
Energetic arousal
Energetic arousal is a non-directional component of motivation in all of these models of affect. It is also a construct that has been found to be of great heuristic importance in theories of motivation and cognition ever since Broadbent (1958). More importantly, in that individuals seem to differ systematically in their level of energetic arousal, it is a way to link theories of individual differences to theories of behavior.Even though Corcoran's (1965) definition of arousal as the "inverse probability of falling asleep" is immediately understandable, it is clear that the use of arousal as a construct is problematic. Indeed, some prefer to avoid discussing arousal and use a broader term, energetics, that subsumes many different constructs of motivational intensity and the effects of many environmental stressors (Hockey, Gaillard & Coles, 1986).
Most simply, arousal is a hypothetical construct used to organize the common behavioral effects of exercise, stimulant drugs, sleep deprivation (negatively), time of day, time on task and impulsivity (Anderson, 1990). Each of these separate variables has both a common and specific effect on behavior. Caffeine and amphetamine both make one more alert and able to respond more rapidly and for longer periods of time. Caffeine differs from amphetamine in the locus of its action (post-synaptically versus synaptically) as well as in some peripheral effects (e.g. caffeine induces hand tremor). It is not difficult to demonstrate that different manipulations of arousal have somewhat different effects on the patterning of responses. As an example of a behavioral dissociation, simple reaction time is facilitated by caffeine but is also faster for high impulsives (thought to be less aroused than low impulsives). High impulsives differ from low impulsives in terms of their speed accuracy tradeoff (Dickman and Meyer, 1988) as well as in terms of arousal level.
Sanders (1983) discussed the multiple approaches to the study of stress. One can manipulate the antecedent conditions or examine the physiological consequences. Similarly, there are at least three ways to study the relationship of arousal to performance: 1) by varying the situational demands thought to lead to arousal; 2) by correlating psychophysiological measures to performance; and 3) by correlating self report measures of arousal with performance.
The first approach, manipulations of arousal by the use of stressors such as stimulant drugs, noise, time on task, or time of day, is more commonly used by experimental psychologists. Broadbent's 1971 review suggested that there were common effects for some of these manipulations, but also showed that at least two levels of control processes needed to be invoked to understand all of the effects. A lower level of control associated with executing well learned responses was thought to be sensitive to noise or sleep deprivation and an upper level control process responsible for monitoring the state of the lower level process was thought to be sensitive to alcohol, extraversion, and time on task.
Hockey (1986) has proposed that each manipulation produces its own idiosyncratic state, and that it is a mistake to look for a holy Grail of unified arousal. Several energetics theorists (Gopher, 1986; Mulder, 1986; Sanders, 1983, 1986) have made use of Pribram and McGuinness' (1975) distinction between (phasic) arousal as affecting input processes, (tonic) activation as affecting motor outputs, and effort as an integrative resource allocation mechanism (Figure 2).
2) Levels of control associated with reaction time. Different stages of
information processing are affected by different control processes. Modified
from Mulder's (1986) revision of Sanders' (1983) model of reaction time.
After reviewing the parallels and differences between physiological measures
and psychological manipulations, Broadbent (1971) concluded that "We have
therefore no satisfactory physiological reference for the general state which
we are discussing, and which we have revealed purely from behavioral studies.
In some ways it might have been better therefore to avoid using the term
'arousal' for this behaviorally defined concept, but this would probably do too
much violence to the common usage in the literature. The reader should
remember however that we are working solely on a psychological level, and that
the existence of a physiological concept of arousal is merely an interesting
parallel, with no direct contact at present." (p 413). Later he added that "in
complicating the theory of arousal we shall need to know more about the
functions involved in various tasks; behavioral studies and the physiological
attack upon the brain must go hand in hand." (p 447).
The second approach, that of psychophysiological correlates, has proven to be
the most difficult. Partly this is due to confusing a within subject concept
with between subjects measurement (Venables, 1984). It is also partly due to
variations in the time course of different physiological measures. Just
as broad motivational constructs affect behavior at different time courses, so
do narrow constructs of motivational intensity, (e.g., arousal) have different
temporal parameters. EEG measures of arousal have latencies measured in
milliseconds while autonomic measures such as Skin Conductance have latencies
measured in seconds, and body temperature reflects average levels of metabolic
demands during the previous several hours. The disassociations and specific
patterning of responses associated with reactions of the hand, the heart and
the head (Lacey, 1967) make physiologists particularly cautious whenever they
discuss a construct such as generalized arousal.
The third approach is to use self reports of arousal. Thayer (1989) has argued
that subjective estimates of energetic arousal are the most likely to be
associated with performance. He has also reported that self ratings correlate
more with psychophysiological measures than the measures do themselves. This
is what would expect if each psychophysiological measure had specific as well
as general effects, and if subjective awareness of arousal reflected the
general effects.
Matthews (1989) and his colleagues (Matthews, et al., 1989) have done some of
the most extensive work examining the relationship between self reported mood
and performance. They have found consistent, although complicated,
relationships between self reports of energetic arousal and performance on a
variety of simple and complex detection tasks. In addition, they have found
that state measures of self reported arousal interact with trait measures of
individual differences in introversion-extraversion to affect performance on
these tasks.
The use of the term arousal to encompass phenomena ranging across many orders
of temporal magnitude from the milliseconds of the early stages of the evoked
potential (Mulder, 1986) to the effects of 10 minute brisk walks (Thayer, 1989)
to the tendency to seek out stimulation throughout a lifetime (Zuckerman, 1991)
is thought by many to be a mistake. I disagree. I believe that the concept
that changes in resource availability are associated with changes in arousal
allows one to integrate the effect on cognitive performance of stable
personality traits with those of variety of environmental manipulations. This
model has great heuristic value, for it allows an integration of seemingly
unrelated phenomena. Such broad lumping together of disparate effects does
indeed mask differences, however. Each task and each measure has its own
unique variance as well as common variance. What is important is to try to
distinguish the unique from the shared variance. But this is the fundamental
challenge of any theory.
Personality,
motivation, and performance
Over the past 17 years, my colleagues and I have examined how personality
traits combine with situational manipulations to produce motivational states
that in turn affect cognitive performance. For organizational purposes, these
effects can be conceived as affecting information processing at several
different, possibly overlapping, stages (Figure 3).[8] The conceptual stage model I present is
obviously derived from Broadbent's filter model (1958) and the latter
distinctions between filtering and pigeonholing (1971), and even more from his
Maltese cross model of memory and attention (1984) as well as Sanders' (1983)
stage model of reaction time. I show it merely to distinguish between the
types of demands placed upon the subject. Stimuli must first be detected, then
encoded, before this new information is able to be stored in memory. Based
upon the incoming stimuli, further information needs to be retrieved from
memory, information needs to integrated, and some response needs to be
executed. This is a continuous loop, in that as a consequence of each
response, environmental feedback occurs that partly determine the next stimulus
that is to be detected. Storage and retrieval processes are shown as arrows
between the encoding, integrating, and memory systems.
3) A conceptual organization of the stages of processing that are affected by individual differences in motivation. Environmental inputs are first detected, then encoded, stored, and integrated with prior expectancies before responses are selected and executed. Behavioral acts, through feedback, lead to new environmental input. Storage and retrieval processes are represented as arrows to and from the memory system.
Motivation affects each of these stages. In terms of tasks we have examined, we believe that vigilance-like tasks relate to the detection and response stages and are affected by variations in arousal; individual differences in the learning of affectively valenced material occur at the encoding stage and are related to differential sensitivities to rewards and punishments; memory storage and retrieval and the effect of retention interval are affected by variations in arousal: arousal facilitates storage but hinders retrieval; and the information integration stage is curvilinearly related to arousal because it reflects two components--a beneficial effect due to the speed of input and a detrimental effect due to unavailability of recent events.
On a larger time scale, as the information processing loop continues to be executed, resources vary in their availability and in their allocation. Knowledge structures in memory change, affective reactions to the outcomes bias expectancies of future reinforcement and strategic decision processes are used. The encoding of environmental demands reflect differences in biological sensitivities to cues for rewards and punishment (Gray, 1981) as well as the prior contents of memory. Emotional reactions to feedback reflect the interaction of expectancies and outcomes. Positive affective states result from reward following expectancies of reward or non-punishment following expectancies of punishment. Negative affective states result from punishment following expectancies of reward and from punishment following expectancies of punishment (Rolls,1990). Positive affect facilitates approach behavior, negative affect facilitates avoidance behavior. Approach and avoidance tendencies are mutually inhibitory. Increased arousal facilitates the detection and storage of information as well as the execution of the dominant response tendency. This leads to a much more complex model (Figure 4), but one that is probably necessary if the interdependent effects of cognitive and affective processes are to be understood. This model is an attempt to sketch out the systems that are involved in actively processing valenced information in an ongoing system responding to environmental demands and environmental reinforcements[9].
4) Affective and cognitive reactions as part of an ongoing behavioral system. The encoding of environmental demands reflect differences in biological sensitivities to cues for rewards and punishment (Gray, 1981) as well as the prior contents of memory. Reactions to feedback reflect the interaction of expectancies and outcomes. Positive affective states result from reward following expectancies of reward or non-punishment following expectancies of punishment. Negative affective states result from punishment following expectancies of reward and from punishment following expectancies of punishment (Rolls,1990). Positive affect facilitates approach behavior, negative affect facilitates avoidance behavior. Approach and avoidance tendencies are mutually inhibitory. Increased arousal facilitates the detection and storage of information as well as the execution of the dominant response. Adapted from Clark and Watson (1991), Gray (1981), Larsen (1991), Rolls (1990), and Thayer (1989).
In the following section I discuss the immediate motivational effects upon performance of various combinations of individual differences in personality and situational stressors. In the final section I suggest how an adequate theory of individual differences and cognitive performance needs to examine motivational effects on the stages of processing as well as consider the larger temporal variations in affect, cognition and behavior that occur as the information processing loop continues over time.
Personality, vigilance and continuous performance
Differences in the ability to sustain performance across time have been noticed in dogs, sonar operators, train engineers, and faculty listening to colloquia. What is particularly interesting to those interested in coherent descriptions of personality is that several of the basic dimensions of personality are related to performance decrements across time.After an extensive discussion demonstrating that performance decrements generalize across several types of continuous performance tasks, Broadbent (1958) presented evidence suggesting that extraverts were more likely to show such decrements than were introverts. By 1971, the evidence supporting this position was much stronger. Extravert performance deteriorates more rapidly in terms of detecting infrequent signals (Bakan, Belton & Toth, 1963; Keister and McLaughlin, 1972), in terms of variability and speed of continuous reaction time (Thackray, Jones & Touchstone, 1974), and in the ability to stay awake on long distance drives (Fagerström & Lisper, 1977).
Matthews (1989) and Matthews, Davies and Lees (1990) have shown that this decrement in performance can occur very rapidly and that self reported high arousal is associated with the ability to maintain performance. They used a rapidly paced discrimination task introduced by Neuchterlein, Parasuraman and Jiang (1983) with an inter stimulus interval (ISI) of 1 second and a priority stimulus frequency of 25% (i.e. a response was required on the average every 4 seconds) with two levels of stimulus degradation. For the degraded stimuli, performance of low aroused subjects deteriorated within 12 minutes, but did not for high aroused subjects. Neither Neuchterlein et al. nor Matthews et al. found a decrement on this task with non-degraded stimuli. Neuchterlein et al. interpret the detection of the degraded stimuli as requiring substantial "effortful" processing as compared to the non-degraded stimuli which may detected in an "automatic fashion". Matthews et al. argue that the degraded stimuli lead to the kind of resource limited attention task that Humphreys and Revelle (1984) suggest should benefit from high arousal.
Impulsivity at the adult level has frequently been claimed to be related to the impulsivity associated with hyperactivity or what has come to be called Attention Deficit Disorder (ADD) with (or without) Hyperactivity. ADD children are particularly susceptible to decrements on continuous performance tasks. Sergeant and van der Meere (1990) have reviewed the application of Sanders' model of energetic effects on reaction time to the case of individual differences associated with attention deficit disorders. Their review is an excellent example of the wealth of information that comes from combining sophisticated experimental procedures with the study of important individual differences.
Revelle, Rosenberg & Anderson (in preparation) have recently completed three studies with an even simpler task than the Neuchterlein et al. task, but one that still shows pronounced decrements within a few minutes. Because of our interest in the dynamics of behavior, we examined performance as a function of time on task. The task we used (variable fore-period reaction time with an inter stimulus interval of 1-11 seconds) lasts for just a few minutes (12-15) and is typical of the demands placed upon subjects doing many monotonous real world (or experimental) tasks. The subjects task is to respond as rapidly as possible whenever a series of X's appears on the monitor of a computer. The targets remain until the subject responds. The fastest reaction times of our subjects tend to be of the order of 220-250 msec, with most responses being less than 400 msec. We discard all trials in which the subject took more than 1000 msecs to respond, although we have observed at least one subject who was taking 7-8 seconds on some trials. That is, our task succeeds in putting some subjects to sleep. More objectively, self reports of energetic arousal decay reliably across the 12 minutes of the task.
We have done three studies with this task. The first examined the effects of impulsivity, anxiety (neuroticism) and time of day (0900 versus 1930 hours) the second added caffeine as a factor; the third was run just in the morning and examined the effects of an incentive (half of the subjects were offered $10 if they could score in the top 33% of the subjects, the other half were not told about the incentive). Dependent measures were simple reaction time, as well as the change in reaction time as a function of trials.
When the results from all three studies are compared they clearly show a difference between the effects of (caffeine induced or diurnally varying) arousal versus (monetary incentive induced) effort. Although both arousal and effort manipulations improve performance, only the arousal manipulation was able to sustain performance. The change across time clearly demonstrated the effects of arousal as well as impulsivity and neuroticism. Impulsivity was positively correlated with decay of RT in the morning but negatively in the evening, and high neurotics were unable to maintain their performance from the first to the last part of the experiment. These results bring to mind Broadbent's (1971) two levels of control. For although effort facilitated reaction time (Broadbent's lower level) arousal facilitated the long term maintenance of reaction time (Broadbent's higher level).
