Psychological R eviewpersonality-project.org/revelle/publications/hr84.pdf · 154 M IC H A E L S . H U M P H R E Y S A N D W IL L IA M R E V E L L E m ance, w hereas at other tim

Psychological ReviewVOLUME 91 NUMBER 2 APRIL 1984

Personality, Motivation, and Performance:A Theory of the Relationship Between

Individual Differences and Information ProcessingMichael S. Humphreys

University of Queensland,St. Lucia, Brisbane

Queensland, Australia

William RevelleNorthwestern University

We introduce a model to relate the personality dimensions of introversion-extra-version, achievement motivation, and anxiety to efficient cognitive performance.We show how these personality dimensions in combination with situational mod-erators (e.g., success, failure, time pressure, incentives, time of day, and stimulantdrugs) affect the motivational constructs of arousal and effort. We propose a generalinformation-processing model that accounts for the systematic effects of these mo-tivational states on certain task components (sustained information transfer andsome aspect of short-term memory). We combine empirical generalizations abouttask components in a structural model and derive testable predictions that differ-entiate alternative motivational hypotheses.

There are two major approaches to the studyof human intellectual performance. The firstfocuses on the effect of personality and indi-vidual differences, and the second attempts to

The order of authorship is arbitrary. This research wassupported in part by Grant MH29209(01-04) from theNational Institute of Mental Health to both authors, inpart by Grant MH29209(05-07) from the National Instituteof Mental Health to William Revelle, and in part by agrant from the University of Queensland to Michael Hum-phreys. Preparation of an earlier draft of the article doneat Oxford University was supported in part by FogartySenior International Fellowship TW00580 from the Na-tional Institutes of Health to William Revelle.

We gratefully acknowledge the suggestions and com-ments made on earlier drafts of this article by K. Anderson,J. Atkinson, D. Birch, D. Broadbent, C. Duncan, H. J.Eysenck, L. Goldberg, J. A. Gray, M. J. Lynch, G. Mandler,J. Onken, T. Rocklin, and B. Underwood. We particularlyappreciate the many hours of discussion and criticismgiven to us by our students. We would also like to thankN. Block, R. Conti, T. Lederer, L. Smith, and R. Shorefor their assistance in preparing the various drafts.

Requests for reprints should be sent to William Revelle,Department of Psychology, Northwestern University,Evanston, Illinois 60201 or to Michael S. Humphreys,Department of Psychology, University of Queensland, St.Lucia, Brisbane, Queensland, Australia 4067.

develop general laws of cognitive psychologyor information processing. Although these twoapproaches rarely are combined, it is difficultto find an example of cognitive performancethat is not better understood by a combinationof both areas. In this article we propose a the-ory that integrates these two fundamentallydifferent paradigms.

We believe that theoretical and empiricalwork on at least three personality dimensionsposes a problem for anyone interested in therelationships of personality to efficient cog-nitive performance. These three areas of per-sonality (introversion-extraversion, I/E;achievement motivation; and anxiety) havebeen shown in a number of studies to be rel-atively independent. Furthermore, research inall three areas has dealt with the situationaland motivational determinants of efficientcognitive performance. What is particularlyinteresting is that each of these three dimen-sions has been shown to have a complex re-lationship to performance. High levels of in-troversion, achievement motivation, or anxietyare sometimes found to lead to better perfor-

Copyright 1984 by the American Psychological Association, Inc.

153

154 MICHAEL S. HUMPHREYS AND WILLIAM REVELLE

mance, whereas at other times they are foundto lead to worse performance than are lowerlevels on these dimensions.

Perhaps even more interesting is that, al-though these dimensions are relatively inde-pendent statistically,1 theoretical discussionsof each of these areas are very similar. Eachdimension is related to an underlying dimen-sion of motivation (sometimes referred to asarousal, sometimes as drive, or sometimessimply as motivation) that, in turn, is thenrelated to performance. In all three cases, mo-tivation is thought to be a primary determinantof performance. Achievement motivation andintroversion-extraversion are thought to havea curvilinear (inverted U) relationship to ef-ficient performance; anxiety is said to have afacilitative effect on easy tasks and a debilitativeeffect on difficult tasks.

In this article we present an integrative the-ory that differs in several ways from the tra-ditional personality-performance models ofI/E, achievement motivation, and anxiety. Inthe traditional model,2 some personality di-mension (e.g., I/E) is thought to combine witha situational manipulation (e.g., placebo vs.caffeine) to produce a motivational state (e.g.,arousal) that is, in turn, curvilinearly relatedto performance on a specific task (e.g., theverbal portion of the Graduate Record Ex-amination). Thus, caffeine would be expectedto facilitate the performance of extraverts buthinder that of introverts. This type of modelhas been used to explain the interactive effectson performance of various personality di-mensions with various situational manipula-tions. However, rather than limit our analysisto one dimension of personality, one situa-tional manipulation, or one performance task,we now integrate the seemingly diverse effectsof different personality dimensions, a varietyof situational manipulations, and a numberof different cognitive performance tasks. Ourapproach also differs from the traditional onein that we go beyond the simple hypothesisthat motivation is curvilinearly related to per-formance by specifying one component of in-formation processing that is facilitated by in-creases in both arousal and effort, and anotherthat is hindered by increases in arousal.

We summarize the interrelationships be-tween personality, motivation, and perfor-mance in terms of a structural model (Figure

1). In this model, we show the effects of thepersonality constructs (impulsivity,3 achieve-ment motivation, and anxiety) in combinationwith situational moderators (e.g., time of day,caffeine, and success and failure feedback) onthe motivational constructs of arousal and on-task effort. These motivational constructs are,in turn, shown as affecting the information-processing constructs of sustained-informationtransfer (SIT) and short-term memory (STM)resources. In keeping with conventional no-tation, we show observed variables as boxesand latent variables or hypothetical constructsas circles. We represent experimental manip-ulations as triangles and interactive effects asXs. The solid lines in the figure representmonotonically positive effects; the dashed linesindicate monotonically negative effects. Al-though this model is expressed in terms of thecausal effects of several latent variables (cf.Bentler, 1980), it is the result of a conceptualrather than an empirical analysis.

In justifying the model outlined in Figure1, we first describe a general approach to hu-

1 The low reliability of Thematic Apperception Tests(TATs) makes it difficult to assess the relationship betweenachievement motivation and any other variable. (See,however, Atkinson, Bongort, & Price, 1977; Reuman,1982.) Furthermore, although many studies report lowcorrelations between anxiety and achievement motivation(see Atkinson, 1974; Atkinson & Birch, 1978) and betweenanxiety and either I/E or impulsivity (H. J. Eysenck, 1981),it is harder to find studies that measure both achievementmotivation and I/E. We believe that this partly reflects thefragmented nature of the field and partly the separatedomains of inquiry. See Revelle and Humphreys (1983)for a detailed discussion of the measurement issues as-sociated with all three personality dimensions.

2 To avoid appearing overly critical of the work of others,the specific example of a naive model we use is taken fromRevelle, Amaral, and Turriff (1976); examples of slightlymore complex models are from Craig, Humphreys, Rock-lin, and Revelle, (1979), who used two dimensions of per-sonality, and from Revelle, Humphreys, Simon, & Oilliland(1980), who used two situational manipulations (time ofday and placebo/caffeine).

3 Our original work in this area was concerned withevaluating H. J. Eysenck's theory of I/E. As we discusslater, recent studies have suggested that many of the arousal-based effects associated with I/E are actually due to thelower order factor of impulsivity. Thus, although we reviewthe I/E literature, we prefer to develop our model in termsof impulsivity. See Revelle, Anderson, and Humphreys (inpress) and Revelle and Humphreys (1983) for a more com-plete discussion of the experimental utility of separatingthe two lower order components of I/E.

PERSONALITY, MOTIVATION, AND PERFORMANCE 155

Figure 1. Conceptual structural model of the effects of personality, situational moderators, and motivationalstates on information processing and cognitive performance. (Solid lines represent positive influences, dashedlines represent negative influences, and the absence of a line indicates no direct influence. Circles representlatent variables, rectangles represent measureable variables, triangles represent experimental manipulations,and Xs show interactive effects.)


man information processing and show howthe motivational constructs of effort andarousal can be incorporated into that ap-proach. We then show how several differentways of manipulating effort and arousal havesimilar effects on SIT tasks and may have dis-similar effects on STM tasks. We then combinethese motivational and information-processingconstructs into a general model of the rela-tionships between motivation and perfor-mance. Finally, we show how this generalmodel can aid in the construction of specifictheories linking these three personality di-mensions (impulsivity, achievement motiva-tion, and anxiety) to cognitive performance.

Definition of TermsBefore proceeding to develop and justify the

specific assumptions in our structural model,it is useful to consider what is meant by someof the terms we are using. Note, however, thatwith our behavioral approach the constructsof effort and arousal are defined by their spec-ification in the model rather than by self-reportor physiological measurement. The definitionsof the constructs offered in this section onlyserve to indicate the domain of phenomenain which we are interested.

Personality TraitsPersonality traits are stable characteristics

of individual differences that may be used todescribe and to explain behavior (cf. Hirsch-berg, 1978). At the purely descriptive level,traits may represent mere counts of the fre-quency of certain behavioral acts (Buss &Craik, 1983). As such, traits are convenientsummaries of consistent behaviors across dif-ferent situations. In addition to such descrip-tive summaries, traits may also be used as ex-planations of behavior. In this latter case, traitsreflect genetic, biological, temperamental, orlearned bases for behavior. We do not wish,however, to make any distinction betweentemperamental traits, learned traits, or cog-nitive abilities. Rather we prefer to view traitsas latent variables associated with the corre-lations between different behaviors. Such be-haviors can be measured by self-reports, ob-servations of others, or by objective test data(viz., Block & Block, 1979).4

The effect of these latent variables can onlybe estimated from the pattern of correlations(or covariances) within and between differentbehavioral domains. Correlations within do-mains are frequently referred to as reliabilitiesor internal consistencies, whereas correlationsbetween domains are known as convergent anddiscriminant validities (Campbell & Fiske,1956).

Personality StatesPersonality states are the result of the com-

bination of traits and situations. Thus the traitof anxiety in combination with the threat ofevaluation leads to the state of anxiety. Indi-viduals with less trait anxiety are less likely tohave high state anxiety in most situations thanare individuals with more trait anxiety. Traitsare predispositions to states. Differences in theamount of a particular trait only result in dif-ferences in state when the appropriate mod-erating stimuli are present. Thus, high traitanxiety results in high state anxiety only whenappropriate eliciting cues are present. In theachievement-motivation literature, a similardistinction is made between motive and mo-tivation. Individuals with a high motive toachieve (a trait) when faced with an achieve-ment situation will have a higher level ofachievement motivation (a state) than will in-dividuals with a lower achievement motive.

Situational ModeratorsSituational moderators are those charac-

teristics of the situation that in combination

4 It is beyond the scope of this article to consider ingreat detail the difference between traits as descriptionsand traits as causes. One way to view this issue, however,is in analogy to the distinction between components andfactors in psychometrics. Components are the weightedsums of observed variables and as such do not cause thevariables, but are only convenient summaries. Factors, onthe other hand, represent the unobservable common orshared variance of variables. Variables are seen as weightedsums of (hypothetical and unobservable) factors. Althoughfactors can be used as explanations, components may onlybe used as summaries. Traits denned by the act frequencyapproach are analogous to components; traits used as casualexplanations are analogous to latent factors. In a structuralmodel, this distinction is captured by the direction of ef-fects. Factors are presumed to be sources of variance inthe observed variables; variables are sources of variancefor components.


with personality traits, evoke particular per-sonality states. Typical examples of such mod-erating variables include (but are not limitedto) type of feedback (i.e., success or failure),monetary incentive, threat of punishment,time of day, stimulant drugs, depressant drugs,heat, cold, and noise.

MotivationMotivation is a hypothetical construct that

has traditionally been used to describe andexplain differences in intensity and directionof behavior. It is the state that results frpm acombination of individual needs and desireswith the stimulus properties of the situation.Although they are frequently combined undera single motivational construct, it is helpfulto distinguish between the constructs of arousaland effort.

Arousal. Arousal is the state of the organ-ism that in everyday terms means alertness,vigor, peppiness, and activation. Arousal maybe thought of as a conceptual dimension rang-ing from extreme drowsiness at the one endto extreme excitement at the other. Arousallevel is the result of internal and external stim-ulation. High levels of arousal are associatedwith high levels of sensory input; low levels ofarousal are associated with low levels of sen-sory input. Specifically, loud noises, brightlights, time pressure, external distractors, andcomplex stimuli all lead to increases in arousal.

Indicants of arousal include both physio-logical and self-report measures. High arousalis characterized physiologically by higher levelsof skin conductance, heart rate, breathing,metabolic activity, plasma levels of free fattyacid, and plasma levels of epinephrine andnorepinephrine. Electrocortically, arousal isassociated with alpha desynchronization andhigher dominant frequencies. In terms of self-report, arousal is indicated by feeling moreactive, alert, and peppy and less sleepy ordrowsy.

A distinction should be made here betweenwithin-subject and between-subject differencesin arousal. Within individuals, increases in ex-ternal stimulation are correlated with increasesin the physiological measures just mentioned.Between subjects, however, the pattern is morecomplex. Individuals differ with respect towhich physiological system responds the most

to external stimulation. That is, some subjectsrespond to increases in noise with an increasein heart rate, whereas others respond with anincrease in breathing rate. As an example, ifone person increases his or her heart rate by10 beats per minute but only increases his orher breathing rate by 1 breath per minute,whereas another person increases his or herheart rate by 1 beat per minute and his or herbreathing rate by 5 breaths per minute, thereseems to be a negative correlation betweenthese two response systems. However, bothpeople increased their breathing and heartrates in both systems; there was only a differ-ence in relative change in the two systems. Inthis example, there was a positive correlationbetween the response in these two systemswithin individuals; between individuals therewas a negative correlation.

It is partly this issue of within-versus-be-tween individual measurement that has led tothe response-specificity/generality argumentbetween Lacey (1967) and Duffy (1962). How-ever, as Lacey (1967) has made clear, evenwithin subjects the picture is not as clear asone would like. Although normally linked, thevarious response systems can be shown to beindependent of each other. He argued that"electrocortical arousal, autonomic arousal,and behavioral arousal may be considered tobe different forms of arousal, each complex initself" (p. 15). That is, they do covary, butthey can be shown to be separate.

Although there are possibly many differentarousal systems or physiological ways of be-coming aroused,5 it is the common behavioral

5 For example, stimulant drugs could have effects eitherpresynaptically or postsynaptically but have similar effectson the overall rate of neurotransmission. More proble-matical for our theory is the fact that in free-runningexperiments, the sleep-wake cycle and the body-temper-ature cycle can be disassociated (Minors & Waterhouse,1981), This finding indicates that temperature and sleep/wake are not responding to the same system albeit withdifferent lags (Humphreys et al., 1980) but are in partresponses to different systems. This result is a challengeto our theory only if different performance tasks adapt atdifferent rates to shifts in these systems (e.g., shift-workor time-zone changes). At this time the evidence for dif-ferential adaption of performance tasks is quite tentative(Folkard & Monk, 1980; Hughes & Folkard, 1976; Monk,Knauth, Folkard, & Rutenfranz, 1978). Even if there werestrong evidence for differential adaptation, this would notshow that sleep/wake- and the body-temperature rhythms


effects with which we are concerned. We findit helpful to think of arousal as a conceptualdimension defined as that factor common tovarious indicants of alertness. We accept thatthere are specific factors associated witharousal manipulations, but feel that generalarousal is theoretically parsimonious as ahigher order construct. The usefulness ofthinking of arousal as a single activational stateis, however, an empirical question. We hopeto show that conceiving of arousal as a generalfactor allows for a broad synthesis of researchfindings from a variety of personality and ex-perimental paradigms.

It is also useful to distinguish betweenarousal at the macro and at the micro levels.At the micro level, arousal may be indexed bypupil dilation, changes in heart rate (both ac-celeration and deceleration), and changes inthe electroencephalogram (EEG; Kahneman,1973). But these measures do not index arousalat the macro level. At this more global level,arousal relates to the general feelings of alert-ness or activation (Thayer, 1967, 1978), bodytemperature (Blake, 1967), and hormonal ex-cretions (Frankenhaeuser, 1975). It is also atthis macro level that caffeine, time of day, andpersonality can be shown to affect performance(Revelle et al., 1980).

This distinction is partly one of duration.Arousal at the micro level has been associatedwith pathway activation (Posner, 1978), whichis a consequence of such experimental ma-nipulations as a warning signal for a reactiontime test. The effects of such activation arevery transitory and are indexed in millisec-onds. Arousal at the macro level, however, isof much longer duration. The manipulationswe consider to be at this level have effects thatare indexed in minutes or in hours. An evenbroader level of analysis that might be relatedto the arousal concept is sometimes referredto as stress (Seyle, 1976). The effects of variousstressors on an organism usually persist fordays or for months. Although it is obvious thatall three levels interrelate, our concern is atthe macro level.

are different forms of arousal. The sleep-wake cycle isnormally associated with the body-temperature cycle andwith arousal. However, when these two rhythms are dis-associated, it is possible that sleep/wake is more associatedwith variations in effort than it is with arousal (Minors& Waterhouse, 1981).

Effort. Effort is the motivational statecommonly understood to mean trying hardor being involved in a task. Effort is increasedwhen the subject tries harder, when there areincentives to perform well, or when the taskis important or difficult. Effort can be appliedto one task rather than another through ex-perimental instructions or changes in the pay-offs (rewards and punishments). A distinctionneeds to be made between the subjective feelingof trying hard and on-task effort, which wedefine later as the allocation of resources tothe task at hand. Our theory is one of on-taskeffort rather than the subjective feeling oftrying hard.

Arousal and effort are both hypotheticalconstructs. We think of these two constructsas separate and link them together under therubric of motivation only to be consistent withprevious usage. We acknowledge that exper-imental manipulations of one may affect theother and that many theorists have combinedthe two constructs. We believe, however, thatit is useful to distinguish between the effectsof stimulant drugs, time of day, and lack ofsleep (arousal), and those of incentives, im-portance, difficulty, and instructions (effort).In some sense, this distinction is one betweenbiological manipulations and cognitive ones.One of the major goals of this article is toprovide a system whereby we can differentiatethe constructs associated with these differentsets of manipulations.

Motivation and Information ProcessingBefore proposing our specific theoretical

ideas, we feel that it is worthwhile to considerin general terms (a) what a theory of the re-lationship between personality and perfor-mance should be and (b) why such a theoryis needed. A theory about personality and per-formance should cover the entire range fromindividual differences and situational moder-ators to information-processing constructs andperformance. Until now, theories of individualdifferences have been related to performancevia ill-defined constructs. For example, a com-mon assumption has been that performanceis curvilinearly related to arousal. As such,this assumption is practically useless. It ismerely a description of data, not an expla-nation. Furthermore, without a specification


of the information-processing components in-volved, it is impossible to predict whether anygiven experimental task should show a cur-vilinear relationship with performance.

We believe that to make progress in thisarea, we need a theory that goes beyond theassumption of curvilinearity. Specifically, sucha theory should get rid of the curvilinear as-sumption and replace it with two or moremonotonic processes, at least one of whichimproves with increases in arousal and at leastone of which deteriorates. These two processes,in combination, can produce curvilinearity.Models of this type have the advantage thatthey specify the tasks or kinds of tasks in whichperformance increments or decrements canbe expected. Such a theory of two monotonicprocesses is also easier to test than is an as-sumption involving curvilinearity.

In keeping with Navon and Gopher (1979),we assume that the human information-pro-cessing system can be described as being com-posed of multiple allocatable resources:The human system is probably not a single-channel mech-anism but rather a complicated system with many units,channels, and facilities. Each may have its own capacity(which is, roughly, the limit on the amount of informationthat can be stored, transmitted, or processed by the channelat a unit of time). Each specific capacity can be sharedby several concurrent processes; thus it constitutes a dis-tributable resource. (Navon & Gopher, 1979, p. 233)

There are two obvious ways to map moti-vational constructs into this general approachto information processing. The first is in termsof the concept of allocation. If a subject isperforming two tasks simultaneously, we canplot performance on Task A as a function ofperformance on Task B. In general, across in-structions and incentives designed to increasethe importance of performing well on Task A,relative to the importance of performing wellon Task B, there is a smooth trade-off betweenperformance on the two tasks. That is, as per-formance on one task improves, performanceon the other task tends to stay the same or togradually decline. There are usually no abruptchanges in performance on either task. Theconclusion that is drawn from such a perfor-mance-operating characteristic curve is thatthe subject is taking resources from one taskand allocating them to the other task.

When only a single task is being performed,it is also possible to think about the allocation

of resources. Basically there are three ways inwhich this could occur: (a) If the task has mul-tiple components, the subject may allocate re-sources from one aspect of the task to another.For example, subjects may sacrifice accuracyin order to increase their speed (Pachella,1974), or sacrifice the quality of their answersin order to increase the number of answersthey produce (Bavelas & Lee, 1978). (b) Al-location of resources could also occur betweenan experimenter-defined task and a subject-defined task. Examples of a subject-definedtask include worrying about performance,daydreaming, or attempting to do well on atask component that is not measured by theexperimenter. Performance on the experi-menter's task stays the same, or improves, asincentives to do well on it are increased andoff-task incentives are decreased. Performancestays the same or gets worse as on-task incen-tives decrease and off-task incentives increase.(c) If there is a cost associated with the allo-cation of resources and if each additional unitof resource allocated to a task produces a de-creasing amount of improvement in perfor-mance, there comes a point where the cost ofadding resources equals the benefit gained bythe improvement in performance (Navon &Gopher, 1979). Thus, it is also possible to thinkabout resource allocation even when there isno detectable trade-off between task compo-nents and no detectable subject-defined task.That is, there may be resources that are notbeing used but that are not allocated unlessthe benefits for doing well on the task are in-creased.

A second way that motivation could bemapped onto this general information-pro-cessing model is in terms of the availabilityof resources. In keeping with Kahneman(1973), increases in motivation might actuallycreate additional resources (make them moreavailable), not just reallocate resources fromone task to another or from a pool of unusedresources to the experimenter-defined task.The process that is most compatible with tra-ditional thinking about arousal (Kahneman,1973) is simply that the total number of re-sources increases. It is also possible that anincrease in arousal could lead to a decreasein the cost associated with allocating these re-sources. One way of thinking about this latteralternative is not that there is a reduction in


the cost, but rather that the cost is deferreduntil a later time.

We should note that resource allocation andresource availability are metaphors and thatthere are many ways in which the functionaldistinction expressed by these metaphors canbe preserved. They do, however, serve a pur-pose in communicating our hypothesis abouta functional distinction between motivationalconstructs.

In trying to determine how many motiva-tional constructs are needed to explain efficientperformance, we start by assuming that effortmanipulations affect the allocation of resourcesto on-task endeavors. By this we mean thatthe mechanism that is responsible for perfor-mance trade-offs in the dual-task situation isalso responsible for some of the performancechanges with effort manipulations in single-task situations. Our questions then are whetheranother motivational construct is required,and, if so, what the differences are in the re-lationships of the three personality dimensionsand the two sets of motivational manipulationsto the two constructs. Even though we realizethat more than two motivational constructsmay eventually be required, we wish to explorefirst the more parsimonious hypothesis that atmost two are required. In the next section wespecify some of the processes that get betterand some of the processes that get worse withmotivation. Once some of these have beenspecified, it is possible to specify the kind ofresults that would discriminate among alter-native hypotheses about the number and thenature of motivational constructs.

Motivation and Performance IncrementsIn our search for processes that improve

with arousal and effort, we have looked at per-formance changes in reaction time, letter can-cellation, vigilance, and simple arithmetictasks as a function of incentives, time of day,sleep deprivation, and the stimulant drugsamphetamine and caffeine (Revelle & Hum-phreys, 1983). The results are by no meansentirely consistent, yet some firm conclusionscan be drawn. Both an increase in incentivesand an increase in arousal can improve speedand/or the number correct on these taskswithout a compensating increase in the errorrate.6 Furthermore, there is almost no evidence

that the stimulant drugs (caffeine and am-phetamine) impair performance on these tasks(also see Gilbert, 1976; Weiss & Laties, 1962).7

The issue about whether overarousal deficitsoccur with noise and heat is not as clear. Poul-ton (1976) has reviewed the literature on theeffects of heat, continuous noise, and inter-mittent noise on a variety of tasks. The studieshe cited clearly show that these manipulationscan improve letter cancellation, reaction time,vigilance, and simple arithmetic, but there arealso many cases (Poulton, 1977) where thesemanipulations hurt these tasks. It is contro-versial, however, as to whether this debilitationis due to overarousal or to distraction, differ-ential transfer, masking, and underarousal(Poulton, 1977, 1978, 1979). The issue is cer-tainly not resolved (see Broadbent, 1978;Hartley, 1981; Poulton, 1981). Nevertheless,because Poulton has provided plausible ex-planations for at least some of the deleteriouseffects of noise and heat on these tasks andbecause comparable deficits do not occur withthe stimulant drugs, we feel that it is worth-while to explore a model of human perfor-mance that asserts that both arousal and effort

6 Many authors did not report error rates, so it was notalways possible to tell if subjects were sacrificing accuracyin favor of speed. Still, enough studies reported error rates, •or in the vigilance tasks reported d's, to ensure that anabsolute improvement in performance can occur.

7 Hollingworth (1912) reported that small doses of caf-feine hurt simple reaction time, whereas larger doses im-proved performance relative to a placebo. This result,however, is certainly not an overarousal effect as it wasonly the intermediate dose that hurt performance. As theauthor suggested, it may have been due to a change incriterion. Wenzel and Rutledge (1962) reported that smalldoses of caffeine (100 and 200 mg) improved complexvisual reaction time relative to a placebo but that a 300-mg dose hurt performance. In the same study large dosesof amphetamine did not similarly hurt performance, sothis finding has to be viewed with some caution. Frowein(1981) has also reported that amphetamine hurt choicereaction time with an incompatible stimulus-responsemapping, but not with a compatible mapping. This findingwas not replicated in a subsequent study (as reported byFrowein, 1981) and must be considered tentative. TheWenzel and Rutledge (1962) study also involved a complexbut not necessarily incompatible mapping, so it is possiblethat the effects of stimulant drugs on reaction time aremoderated by stimulus-response compatability. It is in-teresting to note that incompatible stimulus-responsemappings introduce an episodic-memory component intothe task, so these effects may be examples of memory-induced arousal deficits.


manipulations increase the number of re-sources that are applied to these tasks.

We characterize these tasks (reaction time,vigilance, simple arithmetic, and letter can-cellation) as information transfer (IT) tasks.By this we mean that the subject is requiredto process a stimulus, associate an arbitraryresponse (Posner, Snyder, & Davidson, 1980)to the stimulus, and execute the response. Fur-thermore, there is no appreciable retention ofinformation required nor is there an appre-ciable amount of distraction. In addition,many of the tasks reviewed involve either tem-poral uncertainty or rapid pacing for an ap-preciable length of time. Indeed, a commonobservation was that the effects of arousal ma-nipulations tended to show up only in the latterstages of an experimental session. Our con-clusion, thus, is that both effort and arousalmanipulations improve SIT.

Theoretical predictions. We assume thatboth an increase in on-task effort and an in-crease in arousal lead to an increase in thenumber of resources that are used to sustainIT. Furthermore, performance on vigilance,simple arithmetic, letter cancellation, and re-action time is assumed to be a monotonicallyincreasing function of the number of resourcesapplied. There is a point, however, where fur-ther increases in resources do not lead to in-creases in performance. In Norman and Bob-row's (1975) terminology, this point marks thedata-limited region and performance is limitedby the quality of the external data and thesensitivity of the subject. The region whereperformance does change with changes in re-sources is referred to as the resource-limitedregion. These empirical generalizations aboutimprovements in SIT with increases in on-task effort and arousal along with the standarddistinction between data and resource limi-tations permit us to understand some of theanomalies present in the motivational litera-ture and to make some predictions.

The first issue to be considered concernsthe extremely varied findings in this area. Oneinvestigator may report that stimulant drugsimprove reaction time, whereas another ob-serves no effect. Sometimes investigators reporteffects for some subjects but not for others.For example, Rapoport et al. (1981) reportedthat caffeine improves the performance ofnormal children and leaves the performance

of adults unchanged. Many of these apparentinconsistencies are understandable given theidea that performance can be either resourcelimited or data limited. To illustrate, considerstudies by Frowein (1981), and Sanders,Wijnes, and van Arkel (1982). Frowein (1981)reported that both amphetamine and visualdegradation had significant effects on reactiontime. Amphetamine reduced reaction time andvisual degradation increased reaction time.There was no evidence, however, that thesetwo variables interacted. In a basically similarstudy, Sanders et al. (1982) found that bothsleep deprivation and visual degradation in-creased reaction time, but that these two vari-ables also had interactive effects. That is, theeffects of sleep deprivation were considerablygreater with visually degraded stimuli thanwith visually nondegraded signals. Sanders(1983) has used this differential interactionpattern as evidence for the differential effectsof these three variables on motivational states.He believes that the concept of arousal shouldbe subdivided into two states (arousal and ac-tivation), with sleep deprivation affecting bothstates and amphetamine affecting only acti-vation. Frowein's (1981) failure to find an in-teraction between amphetamine and stimulusdegradation can be easily explained, however,by assuming that a normally rested subject isin the data-limited range with respect to per-formance with these visually degraded stimuli.This hypothesis can be tested if the initialarousal level of the subject can be lowered.Thus, we would predict that there would bean interaction between amphetamine andstimulus degradation for sleep-deprived sub-jects.8

8 Sanders (1983) also recognized the importance oflooking for an interaction between amphetamine andstimulus degradation on sleep-deprived subjects. His po-sition, however, is that such an interaction would be evi-dence for an effect of amphetamine on what he calls ac-tivation as well as on what he and we refer to as effort.Our position, however, is that without evidence for dif-ferential effects of amphetamine and sleep deprivation onvisually degraded signals, Sanders (1983) does not havesufficient evidence to distinguish between activation andarousal. That is, the entire distinction would rest on afailure to find an effect of barbituates on movement time(Frowein, 1981), as compared to finding effects of bothsleep deprivation and amphetamine on movement time(Frowein, Reitsma, & Aquarius, 1981). There are several


Our generalization of this viewpoint is thatstimulant drugs and other arousers are mostlikely to improve SIT performance when atask requires many resources, when the sub-jects are at a low level of arousal (early in themorning or when fatigued), or both. Tasks thatrequire a large number of resources includebut may not be limited to those with temporaluncertainty (Frowein et al., 1981; Sanders etal., 1982), degraded stimuli (Frowein, 1981;Sanders, et al., 1982), and spatial uncertainty(Sanders & Reitsma, 1982). In addition, as asubject acquires experience with a task, fewerresources are required.

When stimulant drugs improve the perfor-mance of children but leave the performanceof adults relatively unchanged (Rapoport etal., 1981), a similar explanation applies. Thereare certainly many differences between chil-dren and adults. They may differ in their levelof arousal, in how hard they try, and in thenumber of resources they have. The mostcompelling difference between children andadults, however, lies in their relative experi-ence. Adults have acquired many skills thatchildren do not yet have and can perform sometasks automatically that children must stillperform under conscious control. The resultis that on many of these tasks the adult requiresfewer resources than the child does in orderto do the task. Thus the adult is more likelyto be in the data-limited range than is thechild. This idea can be tested in two ways:The adults should be affected by the stimulantdrugs (a) if the task was made more complexor (b) if their arousal level was lower.

Resource availability versus data limita-tions. One final issue needs to be addressed.If all of the tasks we have discussed utilize thesame resources, then why aren't there sub-stantial intertask correlations? To explain thelow intercorrelations typically found betweenSIT tasks, it is useful to consider the impli-

reasons why we do not think this contrast is very com-pelling. First, it rests on a null result that needs to bereplicated. Second, there are enough differences betweenthe Frowein (1981) and the Frowein et al. (1981) studies(especially in the average movement times) to question thecomparability of the results. Third, the effects of barbituateson performance are likely to be complex due to the knownanti-anxiety effects (which should lead to an increase ineffort and a decrease in arousal).

cations of task-specific data limitations. InFigure 2 we show a family of hypotheticalcurves relating performance to the number ofresources allocated to a task. Note that thesecurves differ in both the asymptotic levelreached and the rate at which they approachthe data-limited region. That is, some indi-viduals are data limited at a higher level ofperformance than are others and some indi-viduals approach the data-limited region faster(require the use of fewer resources) than doothers. Both of these parameters presumablydepend in a complex way on innate abilities(perceptual acuity, eye-hand coordination,etc.) and on experience (the strategies em-ployed, the number of component skills au-tomatized, etc.). The hypothesis that underparticular conditions subjects will differ interms of the number of resources they haveavailable to sustain IT, does not address thequestion of whether they differ in their asymp-totic levels or in their rate of approach toasymptotes. Thus we can see no basis for as-suming that individuals who have a largenumber of resources (in a particular condition)have high asymptotes or require few resourcesto reach the data-limited range. Furthermore,there is no basis to assume that either asymp-totic performance or the number of resourcesneeded to reach asymptote on a vigilance taskis positively correlated with asymptotic per-formance or the number of resources neededto reach asymptote on a simple arithmetictask. That is, the hypothesis that both vigilanceand simple arithmetic depend on the avail-ability of resources that can be used to sustainIT is not a hypothesis about the asymptoticlevel reached or about the number of resourcesrequired to reach asymptote.

It follows from these observations that whenthe rank ordering across subjects of numberof resources available to sustain IT is sub-stantially similar for two tasks, the rank or-derings in terms of performance may not bevery similar. In particular, the hypothesis thatboth tasks depend on common resources doesnot predict that performance will be positivelycorrelated when performance is in the data-limited range. It is only when performance onthe tasks is in the resource-limited range thatthe hypothesis predicts positive correlations.But even here individual differences in skillsand abilities, other than those needed to sustain


AROUSAL

Figure 2. Presumed effects of arousal on information transfer. (Performance on tasks requiring vigilanceand/or the rapid processing of information is helped by arousal. Such tasks may vary in the amount ofarousal required to perform at any given level. The curves on the left represent tasks with lower IT demandsthan do the curves on the right. Points P and D represent the presumed level of arousal of subjects whowere given a placebo or a stimulant drug. Individual differences in asymptotic level are shown to suggestdifferences in nonmotivational variables such as past training, skills, or ability.)

IT, attenuate the correlations. Thus our pre-diction is that correlations between sustainedIT tasks will be higher when the tasks are inthe resource-limited range than when they arein the data-limited range; tasks that respondto manipulations of arousal or effort shouldbe more likely to intercorrelate than shouldthose tasks that do not respond to such ma-nipulations.

Motivation and Performance Decrements

There are of course a large number of waysto produce performance decrements that havenothing to do with motivation. We have al-ready discussed how noise and heat might bedistracting. In this section, however, we areconcerned with performance decrements thatare associated with the motivational constructsof arousal and effort. Given the information-processing approach that we have outlined,three conceptually distinct types of perfor-mance decrements can occur:

1. When two or more tasks are performedconcurrently, performance on one task maysuffer due to the allocation of resources to theother. This type of performance decrementmay be further subdivided into those situationsin which the experimenter defines both tasksand those in which the subject provides the

second task. The experimenter-defined dual-task situations may be divided further intothose situations where the subject is explicitlyinformed about both tasks, those complextasks that require performance on two or moresubtasks (with or without subject awarenessof this implicit dual-task requirement), andthe incidental-learning paradigms where thesubject is not informed until the recall testabout the dual-task requirements.

2. Prolonged exposure to states of high mo-tivation might lead to a state of lowered mo-tivation (fatigue, sleepiness, etc.). Thus, per-formance decrements that are observed aftera long period of exposure to a motivating con-dition may result from lowered levels of mo-tivation produced by the prolonged exposure.

3. It is possible that heightened motivationdirectly weakens some information-processingprocess or faculty. If a particular paradigm orset of paradigms is unusually sensitive to dec-rements with heightened motivation, somefaculty or process common to these paradigmsmay be affected.

We should point out that it is possible tohave decrements as a result of combinationsof these types of situations. Thus, performancemight deteriorate as a result of resource al-location (Type 1) and fatigue (Type 2), or re-source allocation (Type 1) and the weakening


of a particular resource (Type 3). In the fol-lowing discussion, we briefly review each typeof decrement.

Motivation and Dual-Task PerformanceEasterbrook (1959) reviewed the literature

on the effects of increased emotional arousalin dual-task situations. This, along with a sub-sequent review (Anderson, 1981), clearly in-dicates that under some conditions increasesin motivation are accompanied by decreasesin secondary-task performance. Performanceon the primary task tends to stay the same orto show a slight improvement. These resultsare easily explained if under conditions ofheightened motivation, subjects allocate moreresources to the performance of the primarytask. With SIT tasks in explicit dual-task sit-uations, however, we have not found any ev-idence that clearcut arousal manipulationssuch as time of day, normal sleep as opposedto sleep deprivation, and stimulant drugs pro-duce decrements in secondary-task perfor-mance (see Anderson, 1981). Our interpre-tation then is that performance decrements onsecondary tasks result from changes in incen-tives that lead to changes in resource allocation.These changes in incentives may be obvious,as when payoffs are changed or instructionsto do well on the primary or secondary taskare used. There are also a variety of manip-ulations that probably change both the incen-tives for doing well and arousal levels (e.g.,threat of shock, ego involving or threateninginstructions).

It is also possible that the performance def-icits that have been observed when subjectsare performing a single task are due to theallocation or misallocation of resources. Wehave already mentioned that subjects can al-locate resources to off-task endeavors,thoughts, and so forth. They may worry aboutwhat the experimenter is going to do to them,what is going to happen next period, whetherto do their homework, what their friends aredoing, and so forth.

Many performance decrements also haveoccurred with multicomponent tasks, and herea second explanation seems to be required.These multicomponent tasks can be conceivedof as being composed of simpler tasks thatmust be performed simultaneously or in quick

succession. In these circumstances it is possiblethat under conditions of high motivation theallocation of resources to different subtaskswould be less than optimal. Decrements dueto misallocation should depend greatly on thecharacteristics of the particular task, the in-structions given the subject, and the experi-mental setting. Thus, decrements would behard to replicate, and it might be difficult, ifnot impossible, to predict the circumstancesunder which an overarousal deficit would beobserved.

Lowered Arousal Levels andPerformance Decrements

Poulton (1979) has suggested that perfor-mance decrements due to the aftereffects ofnoise might be due to lowered levels of arousal.Similarly, Revelle et al. (1980) suggested thattheir low-impulsive subjects were fatigued inthe evening because they had been morearoused for more of the day than had theirhigh-impulsive subjects. Cohen (1980) has re-viewed the literature concerning the aftereffectsof anxiety-provoking situations, arousing sit-uations, and situations where effort is ex-pended. There may well be some differencesbetween these various kinds of aftereffects, andnot all of these may result in a state that cansimply be characterized as being one of lowarousal. At this time, however, the data simplydo not permit us to make any finer distinc-tions.

Is Some Information-Processing FacultyWeakened by High Arousal?

The most likely candidates for processes ormechanisms weakened by high arousal arethose that contribute to performance on STMtasks.9 Folkard (1975), Hamilton, Hockey, and

9 The only clear distinction between a STM task andthe IT tasks that we have discussed is in the retentioninterval: Memory tasks require subjects to either maintaininformation in an available state through rehearsal or otherprocesses or retrieve information that has not been attendedto for a short period of time. Although it is true that recallimmediately follows presentation of the last item in amemory-span task, there is a retention interval here aswell because, as some of the items are rehearsed, attendedto, or processed, other items are being ignored. We cannot specify just how long a retention interval is requiredto unambigously qualify a task as a memory task. It may


Rejman (1977), and M. W. Eysenck, (1977)have suggested that high arousal hurts STM.We also have reviewed the effects of time ofday, sleep deprivation, and stimulant drugs ontasks that require the retention of informationfor short periods of time (Humphreys, Lynch,Revelle, & Hall, 1983; Revelle & Humphreys,1983). Performance on such tasks is generallyhurt by higher levels of arousal (Blake, 1967;Folkard, Monk, Bradbury, & Rosenthal, 1977;Hamilton, Wilkinson, & Edwards, 1972). Theonly exceptions to this were the effects of stim-ulant drugs such as amphetamine and caffeineon digit span. In general, digit span seemedto have improved slightly or to have stayed thesame when these drugs were used.

A recent study (Anderson & Revelle, 1983)conducted at Northwestern University, how-ever, showed a clear detrimental effect of caf-feine on a task that included a STM com-ponent. A memory-scanning task was usedwhere subjects had memory sets of differentsizes. There was a significant Set Size (two vs.six) X Drug (caffeine vs. placebo) interaction.That is, subjects were hurt by caffeine onlywith the set size of six. This task had previouslybeen used in a study of adaptation to shiftwork (Folkard, Knauth, Monk, & Rutenfranz,1976) where there had been a positive cor-relation between speed and body temperaturefor a set size of two but a negative correlationfor a set size of six. Thus, the effect of caffeineon this task appears to be similar to that ofcircadian rhythms.

In another study, conducted at the Univer-sity of Queensland, Humphreys used a run-ning-memory-span task to provide the inter-ference in a Brown-Peterson paradigm. Sub-jects were first shown four digits, one digit ata time. Then they were shown 0, 4, 8, or 16consonants. Both digits and consonants wereshown at a 750-ms rate and there was a 250-ms pause between each block of four stimuli.After 0, 4, 8, or 16 consonants had been pre-sented, subjects were cued to recall digits or

also be very important to distinguish between items thatare recalled because they have been maintained in anavailable state and items that have to be retrieved from aless available state (Revelle & Humphreys, 1983; Wickens,Moody, & Dow, 1981). For the most part, however, thetasks that we have been concerned with involve both kindsof items in indeterminate amounts.

Table 1Number of Consonants and Digits Recalled as aFunction of Delay, Impulsivity, and Caffeine

Immediateconsonants Delayed digits

Group Placebo Caffeine Placebo Caffeine

Low impulsive 12.98 10.75 6.72 9.21High impulsive 11.48 12.06 8.77 8.02

to recall the immediately preceding block offour consonants. Subjects were strongly en-couraged to pay attention only to the currentblock of consonants. That is, they were in-structed to stop rehearsing the digits as soonas the first consonant appeared and to stoprehearsing the previous block of consonantsas soon as the first member of the next blockappeared. In this study the most importantcomparison is between immediate consonantrecall (collapsed over the number of prior con-sonants) and delayed digit recall (lags 8 and16). These results are shown in Table 1. Therewas a significant Personality (high vs. low im-pulsive) X Drug (caffeine vs. placebo) X RecallCondition (immediate consonant vs. delayeddigit) interaction, F(\, 42) = 4.61, MSe =57.74, p < .05. In the placebo condition thelow impulsives did better on immediate con-sonant recall and worse on delayed digit recallthan did the high impulsives. In the caffeinecondition the low impulsives did better on de-layed digit recall and worse on immediate con-sonant recall. The obvious conclusion is thatthere were trade-offs in terms of the numberof resources allocated to the digit and con-sonant tasks. Whether caffeine was changingthe relative importance of doing well on thesetwo tasks or whether it interfered with someinformation-processing faculty so that subjectswere forced to reallocate resources, remainsto be answered. If they were being forced toreallocate, it could be due to difficulties theyencountered with immediate consonant recallor with delayed digit recall.

A correlational analysis suggests, however,that the difficulty might have been with delayeddigit recall (the retrieval of information froma less available state as opposed to maintaininginformation in an available state). In the pla-cebo condition, a lag of 4 (digit recall) seemed


to act like a lag of 0. The correlation betweenthe lag of 4 and the lag of 0 was .86 and thecorrelations between combined immediateconsonant recall and recall at lags of 0, 4, 8,and 16 were .27, .48, -.03, and —.32, respec-tively. In the caffeine condition, a lag of 4 ap-peared to act like the longer lags. The corre-lation between the lag of 4 and the lag of 0was .33, and the correlations between com-bined immediate consonant recall and lags 0,4, 8, and 16 were .54, .10, -.13, and -.15,respectively.

Our overall conclusion then is that tasksthat require the retention of information overshort intervals (seconds to tens of seconds) aremore likely than most tasks to show deficitsassociated with overarousal. We also found noevidence in our review that moderate incen-tives hurt performance on STM tasks (Gei-selman, Woodward, & Beatty, 1982; Weiner,1966). M. W. Eysenck (1980) has reported astudy in which very large incentives hurt STMperformance, but this deficit may have beendue to the effects of these large incentives onanxiety. A very tempting inference from theseresults then is that some faculty or processassociated with retention over short intervalsis directly and adversely affected by higharousal, but not by high levels of effort. Thisconclusion, however, must remain tentativeuntil the specific faculty or process is identifiedand until more work is done on the effects ofachievement motivation and incentives onshort-term memory.

THEORY OF THE RELATIONSHIP BETWEENPERSONALITY, MOTIVATION, AND

PERFORMANCE

The two motivational constructs of arousaland effort allow us to relate individual differ-ences in personality to cognitive performanceon a variety of tasks. In this section, beforeconsidering how the personality dimensions ofimpulsivity, anxiety, and achievement moti-vation relate to motivation and, subsequently,to performance, we first summarize our reviewof the motivation-performance literature. Wethen show how curvilinear relationships be-tween motivation and performance can be de-composed into a positive relationship betweenarousal and sustained information transfer and

a negative relationship between arousal andsome function of short-term memory. We thenconsider how effort affects both of these com-ponents of information processing. Finally, wereview how each of the three personality di-mension relates to arousal and effort and sug-gest in turn how these three traits, in com-bination with situational moderators, affectcognitive performance.

Motivation, Sustained Information Transfer,and Short-Term Memory

Arousal and Sustained Information TransferHeightened arousal has a beneficial effect

on performance of tasks that involve SIT. Fur-thermore, we have found very little evidencethat there is any performance decrement onthese tasks that is associated with overarousalper se. That is, the performance decrementsthat have been observed appear due either tospecific effects of the arousal manipulation(e.g., distraction with noise) or to changes inincentives that result in the reallocation of re-sources from one task to another. We assumethen that arousal monotonically increases thenumber of resources available to sustain IT.Furthermore, we assume that tasks that onlyinvolve SIT show monotonic improvementsin performance with increases in the numberof available resources.

Effort and Sustained Information TransferCognitive manipulations such as payoffs and

instructions can increase efficiency in vigilanceand rapid decision-making tasks. These ma-nipulations should result in the reallocationof resources from secondary tasks to primarytasks, from off-task thoughts to on-task en-deavors and possibly from a pool of unusedresources to on-task resources. In terms of ourstructural model, effort variables are assumedto affect SIT. As a final point, although we areasserting that increases in on-task effort andincreases in arousal monotonically improveperformance on SIT tasks, it is not necessarilythe case that the process by which performanceis increased is the same. In deriving our specificpredictions, we do need to assume that theresources made available by arousal and thoseallocated to the experimental task because of


increased on-task effort add together to drivethe task toward the data-limited range.10

Arousal, Effort, and Short-Term Memory

We have shown that STM tasks are morelikely than most tasks to show deficits withheightened arousal. Our assumption is thatsome resources that are involved in retentionover these short intervals are reduced byheightened arousal, not that all such resourcesare reduced. Indeed it seems to us that manyof the resources used in sustaining IT wouldalso be helpful in STM paradigms (cf. Bowyer,Humphreys, & Revelle, 1983). Thus the abilityto pay attention and to encode incoming stim-uli might be enhanced, while at the same timethe ability of a subject to maintain informationin a readily available state through rehearsalor other processes (Baddeley & Hitch, 1974;Craik & Lockhart, 1972), or the ability to re-trieve information that has not been main-tained (Wickens, Moody, & Dow, 1981), mightdecrease. To provide for this possibility we havedrawn arrows in Figure 1 indicating positiveeffects of SIT resources on STM tasks. Nev-ertheless, we are assuming that a mechanismor process involved in STM is monotonicallydecreased by increases in arousal but not byincreases in effort.

Curvilinearity as Derived From OpposingMonotonic Processes

In this section we show how increasing SITcoupled with a decreasing memory componentcould combine to produce curvilinearity inmulticomponent tasks. Consider the STMcomponent that is present in many such tasks.Our assumption is that high arousal sometimesmakes information that is necessary to per-form the task unavailable after a period ofonly a few seconds. Furthermore, we assumethat the probability that this occurs is a mono-tonically increasing function of arousal. As anexample, consider a multiple-choice exami-nation. After attending to and reading thequestion, the subject must remember it whileattending to and reading the answers. Perfor-mance depends on whether the question is stillavailable or can be retrieved when the correctanswer is encountered. Obviously, the prob-

ability that the question will be remembereddepends on many factors, including the lengthof the question, the particular material in-volved, and the time interval between readingthe question and looking at the correct answer.In Figure 3 we show this hypothetical rela-tionship between arousal and the probabilitythat information that has been encoded willbe remembered when needed. Two curves areshown that represent different memory re-quirements. Both curves start at 1.0 and de-cline to 0.0, with the curve for the more dif-ficult memory requirements declining morerapidly.

For expositional purposes, all of the com-ponents of the task that require SIT aregrouped together into one hypothetical SITfactor. Figure 2 shows what performance onthis SIT factor as a function of arousal wouldbe if we could somehow measure it in isolation.We have already grouped the memory com-ponents together in our memory construct.The relationship between this construct andarousal has been depicted in Figure 3. Thereis no a priori way to specify an appropriatecombination rule for the SIT and STM factorsin determining performance. However, forheuristic reasons we have used the product of

10 To illustrate this point, consider the effects of am-phetamine on simple reaction time. They are greater withvariable foreperiods than with fixed foreperiods (Froweinet al., 1981; Trumbo & Gaillard, 1975), and it looks asif the main difference between variable and fixed fore-periods is in the extent to which the subject can preparefor the go signal (Niemi & Naatanen, 1981), Thus, subjectscan be highly prepared with a fixed foreperiod in whichthey can anticipate the go signal, but because preparationis thought to be difficult if not adversive (Gottsdanker,1975; Naatanen, 1972), a subject cannot maintain a stateof high preparation throughout a variable foreperiod. If,for this task, we equate on-task effort with preparation(that is, if we assume that effort manipulations increasethe amount of preparation and/or the number of occasionson which a subject is prepared), we can see how effort andarousal might interact. One possibility is that with higharousal, subjects can simply maintain a state of high prep-aration for a longer period of time. (Perhaps the cost ofallocating resources is reduced or deferred.) Alternatively,high arousal might produce fast reaction times even whenthe subject is not highly prepared. The latter alternativereceives some support from the interaction between stim-ulus intensity and fixed-versus-variable foreperiods (Niemi& Naatanen, 1981) as it is possible that high arousal makesweak go signals act like strong ones.


m̂< .600

I\\ HIGH

\LOW

3 4AROUSAL

Figure 3. Presumed effects of arousal on short-term memory processes. (The probability of immediateretrieval is reduced by increases in arousal. Tasks may vary as to how sensitive the memory component isto arousal. The curve on the left represents a task that is more sensitive (has more memory demands) toarousal than does the curve on the right.)

the two components as an indicator of therelationship between performance on a mul-ticomponent task and arousal. This relation-ship is given as the solid line in Figure 4. Wecan talk about tasks as being either transmis-sion limited or memory limited. On the leftside of the graph, performance is limited bythe amount of resources needed to sustain IT.On the right side, performance is limited bythe availability of some of the resources neededto remember information over short intervals.

The relationship between effort and per-formance is depicted in Figure 5. Increasedeffort is assumed to raise the SIT curve butnot to decrease the memory curve. The lowerof the two solid lines thus represents the re-lationship between performance and arousalwith a low level of effort. The higher of thetwo solid lines represents this relationship fora higher level of effort. The result of the higherlevel of effort is to make it very difficult toshow performance increments with increases

AROUSALFigure 4. Curvilinearity derived from two opposing monotonic processes. (The ascending and descendingcurves represent the presumed effects of arousal on information transfer and memory, respectively. Thesolid curve represents the resultant relationship between arousal and complex performance. Performanceto the left of point TM is said to be transmission limited; performance to the right of point TM is said tobe memory limited.)


1.000

ICKOLU

.800 - •

.600 •

.400 • •

.200 • •

0.000

1.000

• • .800

- • .600

- • .200

QiLdLu

.400 <

DCOU-z

0.000

AROUSALFigure 5. Curvilinearity derived from two opposing monotonic processes. (The effects of effort are to improvethe information-transfer resource but not to affect the memory resource.)

in arousal. However, with high effort, peakperformance and performance decrementsboth occur at lower arousal levels.

Summary of Motivation and PerformanceRelationships

We have shown how a general model thatdescribes the effects of arousal and effort uponperformance can account for three differentrelationships between arousal and perfor-mance: monotonically increasing, monoton-ically decreasing, and curvilinear (inverted U).We have suggested that on-task effort andarousal both increase (although perhaps in dif-ferent ways) the resources allocated or availablefor SIT tasks. We also have suggested thatarousal, but not effort, reduces the resourcesavailable for STM tasks. In the following sec-tion we show how important dimensions ofindividual differences may be related to themotivational constructs of arousal and effortand, subsequently, to performance.

Dimensions of PersonalityIn this section we briefly summarize the

theoretical and empirical bases of each of three

personality traits and then relate each of themto the motivational constructs of arousal andeffort. We start first with introversion-extra-version, which we believe is related to arousallevel; continue with achievement motivation,which may be related only to effort; and thenconclude with a discussion of anxiety, whichwe believe is related to both arousal and effort.

Introversion-Extroversion and ArousalIntroversion-extraversion is one of the few

personality dimensions that most personalitytheorists agree is robust enough to identifyfrom study to study and investigator to in-vestigator. The I/E dimension has been iden-tified in behavioral measures (H. J. Eysenck,1947), peer ratings, and self-report inventories(Cattell, 1957, 1973; Howarth, 1976; Norman,1969). It has also been found to be a prominentfactor in a set of items sampled from all ofthe major personality inventories (Browne &Howarth, 1977). It has been reported that dif-ferences in I/E are related to differences inphysiological arousal, vigilance performance,social interaction, sexual behavior, creativity,effectiveness of cognitive processing, suscep-tibility to stress, and many other diverse ex-


perimental and observational findings (H. J.Eysenck, 1981; Lynn, 1981). Perhaps it is mostimportant that individual differences in I/Ehave shown impressive stabilities across severaldecades (Conley, in press).

Compatible with all of these findings isH. J. Eysenck's (1967, 1976, 1981) suggestionthat the chief difference between introverts andextraverts is a difference in basal arousal. Thatis, introverts are more aroused than extraverts.This hypothesis, in conjunction with the beliefthat arousal is curvilinearly related to perfor-mance (Hebb, 1955;Yerkes&Dodson, 1908),predicts that the performance of introverts andextraverts should be differentially affected bymanipulations of arousal. Confirming evidencemay be found in recent studies that have shownthat the cognitive performance of introverts ishurt and that of extraverts is helped by stim-ulant drugs such as amphetamine (Gupta,1977), and caffeine (Gilliland, 1980; Revelleel al., 1976; Revelle et al., 1980).

Introveraion-Extraversion Versus Impulsivity

It is likely, although still somewhat contro-versial, that the impulsivity component of I/E is more related to individual differences inarousal than is the sociability component. Re-velle et al. (1980) found that impulsivity hadmore systematic interactions with both timeof day (which presumably is related to arousaldifferences in the diurnal rhythm) and caffeinethan did either I/E by itself or the other chiefcomponent of I/E, sociability. Their findingswere congruent with those that have shownthat many of the results atributable to I/E areactually impulsivity effects (Amelang & Breit,1983; Campbell, 1983; H. J. Eysenck & Levey,1972; M. W. Eysenck & Folkard, 1980; Gray,1972; Loo, 1979). This distinction between I/E and the lower order factor of impulsivity isimportant in that recent psychometric changesin the measurement of I/E (i.e., the EysenckPersonality Questionnaire, H. J. Eysenck &S. B. G. Eysenck, 1975) seem to have down-played the importance of impulsivity (Claridge,1981; Rocklin & Revelle, 1981), even as theexperimental evidence for its importance hasbeen accumulating.

In summary, the dimension of I/E, or atleast the impulsivity component of this di-

mension, has been shown to interact consis-tently with a variety of arousal manipulations.Performance of high and of low impulsivesdiffers as a function of both time of day andstimulant drugs. We conclude from this patternof results that the impulsivity dimension isarousal related. More specifically, we assumethat equal levels of external stimulation (e.g.,noise, caffeine, time stress) lead to differentlevels of internal stimulation (arousal) for dif-ferent people. That is, for the same dose ofexternal stimulation in the morning, low im-pulsives are more aroused than are high im-pulsives. ''

Impulsivity and Performance

To apply our motivation-performancemodel to the case of impulsivity, we do notneed to assume that the arousal-performancecurve differs for these two groups, but ratherthat the stimulation-arousal relationship leadsto higher arousal for the low impulsives in themorning. Increases in stimulation lead to in-creases in arousal for both groups; similarly,decreases in stimulation or repetitive stimu-lation lead to decreases in arousal for bothgroups. Rather than present a single arousal-performance curve with the low impulsives

1' An alternative assumption is associated with the con-cept of transmarginal inhibition (TMI), which suggeststhat high levels of stimuli evoke inhibitory processes and,as such, are actually de-arousing. This hypothesis thatarousal is curvilinear with external stimulation and thatintroverts without stimulation are at their peak of internalarousal is consistent with the skin conductance data ofSmith, Wilson, and Jones (1983). Smith et al. report thatalthough introverts have a higher basal skin conductancewith a placebo or low doses of caffeine than do extraverts,this difference disappears at high doses and extraverts havea slightly higher level of skin conductance. Unfortunately,they do not report their data in terms of impulsivity butrather use the Extraversion scale of the Eysenck PersonalityInventory, which mixes these two primary factors to anunspecified amount.

Although it is probably the case that arousal does notincrease at extreme levels of stimulation, we find TMI tobe a poor explanation for our performance effects. Spe-cifically, Anderson and Revelle (1982) found interactionsof memory load, impulsivity, and caffeine on a proofreadingtask with a within-subjccts design. As we have discussedin more detail (Revelle et al., in press), it is hard to seehow the TMI hypothesis can account for improvementsand decrements on tasks when memory load varies withinsubjects from item to item.


STIMULATION

Figure 6. Impulsivity, arousal, and complex performance in the morning. (Low impulsives perform betterat lower levels of stimulation than do high impulsives.)

farther to the right, we prefer to present twostimulation-performance curves, one for thelow impulsives and one for the high impulsives.Thus, the classic Eysenckian theory appearsin Figure 6.

At least in the morning, low impulsives aremore aroused than are high impulsives. Thus,in the morning, low impulsives should havemore resources for SIT and less of some STMresources. If we assume that the arousal levelof the low impulsive is near the memory-lim-ited range for complex tasks, then increasesin arousal and the resultant decrease in someSTM resources lead to a decrease in perfor-mance. For the high impulsive, on the otherhand, performance is initially resource limitedby the lack of SIT resources, and increases inarousal should lead to increases in perfor-mance (Figure 7). By evening, however, thelow impulsives are now less aroused than arethe high impulsives, and the effects of caffeineare reversed. Caffeine facilitates the perfor-mance of the low impulsives and hinders thatof the high impulsives.

An interesting prediction that follows di-rectly from the model, and one which we donot believe has been studied, is that effortshould have a differential effect on high andon low impulsives. Low impulsives, normallybeing more aroused, are closer to the data-limited region of their SIT performance curve.High impulsives, on the other hand, becauseof their lower levels of arousal, are more able

to benefit from effort manipulations that in-crease their number of SIT resources (referahead to Figure 7). This further suggests thatlow impulsives, whose behavior is not as af-fected by effort manipulations as that of thehigh impulsives, should not learn to be as re-sponsive to social cues as high impulsives. Thisleads to the speculation that perhaps the re-lationship between impulsivity and sociabilityis this hypothesized greater responsivity of thehigh impulsive to incentive motivations in themorning.12

These predictions are in striking contrastto the dimension of achievement motivation,which (in our terminology) has never clearlybeen related to arousal but rather has beenrelated to those situational conditions that af-fect effort.

12 It is tempting to extend this speculation even furtherand to suggest that the primary characteristic of high im-pulsives is a rapid habituation to stimuli. Although theperformance of high and low impulsives frequently doesnot differ early in a task, high impulsives seem unable tosustain their performance (cf. Bowyer et al., 1983). It isas if the arousal level of the high impulsive declines at afaster rate when in a repetitive environment than doesthat of the low impulsive. One of the findings of Revelleet al. (1980) was that the performance of the high impulsiveswas facilitated by caffeine in the evening of the secondday in multiple-day studies. This could be explained bythe decreased novelty (and subsequent de-arousal; cf. Gale,1977) of the same experimental task given the previousday for the high impulsives.


1.000

IoLJ

.800 - •

.600--

.400 -

.200--

0.000

/ \ / \/ LOW \ / HIGH \

- • .800

--.600

1.000

STIMULATIONFigure 7. Impulsivity, arousal, and complex performance in the morning. (Low impulsives are in the memory-limited portion of the curve and experience decrements in performance with increases in arousal. Highimpulsives are in the transmission-limited portion of the curve and experience increments in performancewith increases in arousal.)

QiUb_COz

g.400 <

QioLL.

• • .200

0.000

Achievement Motivation and EffortThe study of achievement motivation is the

study of the needto accomplish something difficult. To master, manipulateor organize physical objects, human beings, or ideas. Todo this as rapidly, and as independently as possible. Toovercome obstacles and attain a high standard. To excelone's self. To rival and surpass others. To increase self-regard by the successful exercise of talent. . . . [It is thestudy of how people] make intense, prolonged and repeatedefforts to accomplish something difficult . . . [and] workwith singleness of purpose towards a high and distant goal.To have the determination to win. To try to do everythingwell. To be stimulated to excel by the presence of others,to enjoy competition. To exert will power; to overcomeboredom and fatigue. (Murray, 1938, p. 164)

Traditional Theories of AchievementMotivation

There are at least three basic approaches tothe study of achievement motivation. The firstis the broad societal theory of David Mc-Clelland (1961); the second is the formal theoryof risk preference of John Atkinson (1957,1964, 1974); and the third is the attributionalformulation of Bernard Weiner (1972, 1978).Of these three approaches, the one that has

been most often applied to the question ofefficient cognitive performance is that of At-kinson and his associates (Atkinson & Birch,1978; Atkinson & Raynor, 1974).

We recently reviewed these relationshipsbetween achievement motivation and perfor-mance (Revelle & Humphreys, 1983) andconcluded that, although achievement moti-vation is difficult to measure, there are someconsistent findings. Specifically, we found thatthe performance of high achievers is betterthan that of low achievers on tasks that are ofmoderate difficulty and that subjects believeare related to ability. Furthermore, the ma-nipulations of achievement motivation are instriking contrast to the ones used in studiesof impulsivity. That is, achievement motiva-tion is affected by telling subjects that the taskis important, is difficult, or is related to long-term goals. Because these manipulations seemvery different from giving stimulant drugs orvarying the time of day, we believe it is usefulto think of achievement-motivation effects interms of how hard subjects are trying to do atask (on-task effort) rather than in terms ofhow alert they are (arousal). We represent thisin our model by a path from achievement mo-


tivation to on-task effort. A distinction shouldbe made here between achievement motivationand approach motivation. We view achieve-ment motivation as one source of approachmotivation but recognize that other sources(e.g., incentives) exist.

An important question is whether there isa relationship between achievement motiva-tion and arousal. Such a relationship is com-patible with the intuitive feeling that engagingin a challenging task leads to an increase inalertness. In this case achievement motivationwould be associated positively with approachmotivation (and subsequently to on-task effort)as well as to arousal. Although we suspect thatachievement motivation relates only to effort,we know of no experiments that allow us toreject the possibility of a path from achieve-ment motivation to arousal. In the next sectionwe derive the predicted consequences of bothof these alternatives and specify the types ofdata that permit us to choose between them.13

Achievement Motivation and PerformanceOne of the advantages of our going beyond

the simple assumption of a curvilinear rela-tionship between motivation and performanceis the ability to distinguish between alternativemotivational models. For achievement moti-vation, the critical data involve tasks with ahigh memory load because both models pre-dict that achievement motivation should bemonotonically related to effort and thus toperformance on SIT tasks.

Model 1: Achievement motivation affectsonly on-task effort. This is the simpler of thetwo models and equates achievement moti-vation with the directional rather than the in-tensity components of motivation.14 What isparticularly interesting about this case is thathigh levels of effort can increase the likelihoodof an overmotivation effect, even though in-creases in effort per se do not lead to decreasedperformance. To understand this, we need toexamine the effect of effort for subjects dif-fering in arousal. Consider first an individualwith a low level of arousal. Increases in effortshould be monotonically related to the allo-cation of SIT resources to the experimenter-defined task, and performance on SIT tasksshould improve until the data-limited rangeis achieved. If the task demands some STM,

then it is possible that this individual will bememory limited before achieving the SITasymptotic (data limited) level of performance.That is, even though the data-limited regionhas not been reached, further increases in re-sources for SIT do not improve performancebecause of the deficit in STM resources. Ad-ditional effort does not hinder performance;it is nugatory (Figure 8).

But now consider an individual with a highlevel of arousal. Once again increases in on-task effort should be monotonically related tothe allocation of SIT resources to the exper-iment-defined task. Such a person, however,reaches the memory-limited region sooner andperforms at a lower level than does the firstindividual. Furthermore, because of this STMlimitation, performance is unaffected by agreat range of variations in effort. Large in-creases in effort result in trivial increases inperformance. Thus, even if achievement mo-tivation only effects effort, if a highly motivatedsubject is challenged with an arousal manip-ulation such as public feedback or time pres-sure, and if the task has a STM component,the result is very likely to be a performancedecrement. This prediction results, of course,in the type of finding reviewed by Atkinson(1974). In this model, high achievement mo-tivation per se does not result in overmoti-vation, but it makes the subject very suscep-tible to arousal-induced deficits. This state-ment can be generalized to any manipulation

13 Yet another possibility to consider is a path from on-task effort to arousal. Many of the predictions of thismodel are the same as those of Model 2, with the exceptionthat any increase of effort (e.g., through incentives) shouldreduce STM resources and lead to deficits on STM tasks.We have already reviewed the evidence against this path(Geiselman et al., 1982; Weiner, 1966) and do not considerit in more detail in this section.

14 We differ here from Thomas (1983) who associateseffort with intensity. We agree with him that persistenceon a task can be used to index effort. We disagree, however,with his equating persistence with intensity. We prefer tothink of time spent on a task as indicating the relativeallocation of resources between different tasks but not thetotal availability of resources. In this context, much of thedata supporting the theory of achievement motivation (At-kinson, 1974; Kuhl & Blankenship, 1979; Revelle & Mi-chaels, 1976) show consistent effects for preferences be-tween tasks or time spent on tasks. Efforts to show dif-ferences in intensity of performance (e.g., Rocklin, 1981)are much less successful.


1.000 .000

0.000 0.000

STIMULATIONFigure 8. Achievement motivation, effort, arousal, and performance: Model 1. (The effects of arousal onsubjects high and low in achievement motivation lead to decrements at lower levels of arousal for high-rather than for low-achievement-motivated subjects. Absolute performance of the high achievers alwaysexceeds or equals that of the low achievers.)

of on-task effort. To us, high on-task effortdoes not lead to decrements in performance,but high on-task effort can not preventarousal-induced (and memory-mediated) dec-rements in performance on tasks that requiresome STM.

What is particularly interesting is the patternof results predicted by this model. Althoughhigh-achievement-motivated subjects exhibitdecrements in their performance when theyare also highly aroused, their performanceshould still be superior to that of less achieve-ment-motivated subjects. That is, the peak ofperformance should occur at lower levels ofarousal for high than for low achievers, butthe absolute level of performance should neverbe less for the high achievers than for the oth-erwise-equivalent low achievers.

Model 2: Achievement motivation affects ef-fort and arousal. Unlike Model 1, which pre-dicts that the performance of high achieversis always superior to that of low achievers,Model 2 predicts that in an arousing situation,the performance of low achievers can exceedthat of high achievers. This follows, of course,from the negative effects of arousal on STMresources. High achievers should have elevated

SIT performance but depressed STM perfor-mance. In Model 1, on the other hand, higherachievers should have elevated SIT perfor-mance but equivalent STM performance. Thepredicted pattern of results for Model 2 areseen in Figure 9.

Both of these models of the effects ofachievement motivation predict that highachievers should do better than low achieverswhen in a nonarousing situation or when per-forming a task that mainly requires SIT re-sources. They both predict that high achieversshould achieve their optimal performance oncomplex tasks (those with some STM require-ments) at lower levels of arousal than shouldlow achievers. They make different predictions,however, for the case of performance in arous-ing situations. Model 1 predicts that the per-formance of high achievers is always superiorto or equal to that of low achievers, whereasModel 2 suggests that the performance of highachievers should fall below that of low achiev-ers. Given the data currently available, we cannot choose between these two models. To dis-tinguish them, it is necessary to show arousal-induced decrements in performance for sub-jects differing in achievement motivation.


1.000

m

I

.800 •

.600-'

.400-

.200

0.000

> / LOW \

- • ,800

- ,600

1.000

crUJCO

o;

g.400 <

ULo

- - .200

0.000

STIMULATION

Figure 9. Achievement motivation, effort, arousal, and performance: Model 2. (The effects of arousal onsubjects high and low in achievement motivation lead to decrements at lower levels of arousal for high-rather than for low-achievement-motivated subjects. At high levels of arousal, the performance of lowachievers exceeds that of high achievers.)

Anxiety, Effort, and Arousal

Although achievment motivation and im-pulsivity seem to be related to different mo-tivational constructs, effort and arousal, in-dividual differences in anxiety have beenshown to be related to both. Anxiety is oneof the most commonly accepted causes of mo-tivationally induced deficits in performance.15

For many years, the arousal/drive theory ofanxiety was predominant (Duffy, 1962; Spence& Spence, 1966). Recently, however, there hasbeen a reinterpretation of anxiety in terms ofcognitive attributions and the direction of at-tention (Mandler, 1975;Sarason, 1975;Weiner,1972; Wine, 1971). Briefly, in this latter in-terpretation, anxiety is seen as a distractor.That is, anxious subjects devote cognitive ca-pacity to worrying about their performanceand thus have less capacity to devote to thetask. A reconciliation of the drive and attri-butional theories can be found in the two-factor anxiety theories of Morris and Liebert(1970) and Schalling (1978). In those theories,some anxiety inductions are seen as affectingprimarily somatic anxiety (arousal), others as

affecting cognitive anxiety (reducing on taskeffort), and others as affecting both. It is in-teresting that this same dichotomy betweeneffort and arousal may be seen in rats, whereanxiety is thought to increase arousal but tolead to behavioral inhibition (Gray, 1982).

As we have shown in our recent review (Re-velle & Humphreys, 1983), the effects of anx-iety on cognitive performance depend oncharacteristics of the task and of the situation.Anxiety can either facilitate or hinder perfor-mance. High anxiety facilitates performanceon easy tasks or when the feedback is positive.High anxiety hinders performance on difficulttasks or when the feedback is negative. Thereis some evidence that the relative contributionsof the cognitive (worry) and somatic (arousal)components change as tasks continue.

15 In the dimensional theory of H. J. Eysenck, anxietyloads very highly on the neuroticism dimension. Althoughwe prefer to discuss the motivational effects of anxiety,certain studies only report results for neuroticism. When-ever this occurs, we assume that the results hold equallywell for anxiety.


Anxiety and Avoidance MotivationOne way to capture this distinction between

the arousal and the effort effects of anxiety isto introduce the concept of avoidance moti-vation. Just as we distinguish between achieve-ment and approach motivation, so it is possibleto separate anxiety from avoidance motivation.Increases in anxiety lead to increases in arousalas well as to increases in avoidance motivation.It is possible, however, to increase avoidancemotivation without increasing either anxietyor arousal (particularly greasy plates increasethe avoidance motivation for washing thedishes, but do not make most people anxious).Avoidance motivation, in turn, reduces on-task effort.

Anxiety and PerformanceIf we equate the worry or cognitive com-

ponent of anxiety with avoidance motivationand a subsequent decrease in resources allo-cated to the task (which we define as a re-duction in on-task effort), and if we allow thesomatic tension (drive) aspects of anxiety torelate to arousal, then it is easy to see howanxiety should have a complex relationship toperformance.

For SIT tasks with a low memory load, wemake two predictions. If the worry componentis in fact a transitory effect,16 then early in thetask, high anxiety should lead to low perfor-mance due to reduced on-task effort. This re-duction in on-task effort is the consequenceof too many off-task thoughts devoted to self-appraisal and negative self-statements. Theadditional arousal induced by anxiety shouldnot improve SIT enough to compensate forthis decrease in effort (Figure 10). As theavoidance component becomes less salient, theanxious subject should improve, first, doingas well as the less anxious subject (the negativeeffect of worry just canceling the positive effectsof arousal) and, finally, doing better than theless anxious one (as the negative worries be-come progressively less important; Figure 11).

For tasks with a high memory load, however,we would expect that high anxiety would havethe equivalent effect of any arouser and leadto decrements in performance. If the task isa complex learning task, then the anxious sub-ject should be initially worse, but should be-come better as the task is restructured to re-

duce the memory load. If the task can be re-structured enough to make the SIT componentmore important, then the high-anxious subjectshould finally do better than the low-anxiousindividual (e.g., Spence, Farber, & McFann,1956).

In a complex task with a high memory load,success and failure feedback, by constantly in-stigating the worry component, should lead toperformance decrements with failure feedbackand increments with success feedback (e.g.,Weiner & Schneider, 1971).

Evaluation of the ModelIn deriving this model we have made some

assumptions that are probably not testable.These include the assumptions that (a) re-sources are limited, (b) they can be sharedbetween two or more tasks, (c) curvilinearity,when it occurs, can be derived from the op-posing actions of two or more monotonic pro-cesses, and (d) trade-offs in dual task para-digms can serve as a model for some of themotivational effects in single-task paradigms.

From these assumptions, we have derivedboth a general model of the effects of moti-vation on performance and a specific modelof how three personality traits relate to mo-tivation and, subsequently, to performance. Inthe general model, we have proposed that atleast two motivational constructs and two typesof information-processing resources are re-quired to explain the data that are availablecurrently. In the specific model, we have sug-gested how individual differences in impulsiv-ity, achievement motivation, and anxiety affectarousal, on-task effort, and cognitive perfor-mance.

The specific model is our interpretation ofhow the theories of I/E, achievement moti-vation, and anxiety may be related to each

16 We use the dynamics of action model (Atkinson &Birch, 1970; see also Kuhl & Blankenship, 1979) to accountfor this transitory effect. If anxiety is seen as a source of"negaction" (avoidance motivation in our terms), then itsinhibitory effect is negatively accelerated because of thecost of resisting the action. Achievement motivation, asan instigating force of the action tendency (approach mo-tivation) on the other hand, instigates the resultant mo-tivational tendency (on-task effort) until the task is initiated.In this model, the effects of inhibitory forces (anxiety) aremost noticeable at the beginning of a session.


1.000

.800 -

CD< .600

I>-

OLJ

.400--

.200-

0.000

STIMULATION

1.000

- • .800

- • .600

• • .400

- • .200

LUU_

Z

I—Zo

QiO

0.000

Figure 10. Anxiety and performance early in the task. (Anxiety is presumed to increase arousal but todecrease effort. This leads to a reduction in the memory resource and to a temporary reduction in thetransmission resource.)

other, to motivation, and to performance. Al- lower levels, high achievement motivation isternative interpretations are possible. For ex- associated with higher arousal as well as withample, it is possible that when compared to greater effort. As we have pointed out already,

1.000

It£O

.800 • •

.600 - •

.400 •

.200 • •

0.000

HIGH \ / LOW \\ \/ ^- -^ \

STIMULATION

• .800

• .600

1.000

LUU_toZ

0.000

Figure 11. Anxiety and performance late in the task. (Anxiety is presumed to increase arousal but not todecrease effort. The information-transfer resource increases as the worry component of anxiety diminishes.)


these alternative specifications lead to slightlydifferent predicted patterns of results and canbe distinguished given appropriate data.

Operating at the general level, we have con-cluded that resource allocation (on-task effort)is not a sufficient explanation for all motiva-tional effects and that the concept of resourceavailability (arousal) needs to be added. Thereare too many instances where performanceimproves with no detectable trade-offs and noindication that incentives for performing wellhave changed. We recognize that it is possiblethat for some situations and/or for some in-dividuals, manipulations that increase effortalso increase arousal (e.g., achievement mo-tivation might be related to both effort andarousal). However, we believe that there aresome manipulations that can change effort butnot arousal. This represents the feeling thatone can try hard without becoming aroused.For example, reading a difficult article requiresa great deal of effort but is not normallythought of as arousing. Similarly, we believethat it is possible to change arousal withoutchanging effort.

The best way to test this distinction is toidentify the information-processing facultythat is hindered by arousal. We have suggestedthat arousal and not effort adversely affectssome aspects of STM. The available data sup-port this conjecture, but we still need directcomparisons of the effects of arousal and effortmanipulations on performance in a variety ofSTM tasks. More generally, the effort/arousaldistinction is testable if we assume that botheffort and arousal increase SIT but that onlyarousal decreases an unspecified information-processing faculty. Then, given two levels ofincentive (/2 > I\) and two levels of arousal(A2 > AI), the combination of 72 and A2 canproduce worse performances than either 72 andA\ or /] and A\, but it can not be worse thanthe combination of I{ and A2.

The next issue for the general model iswhether more than one kind of arousal systemis needed to account for behavior. Debate overthis issue has been clouded by several mis-understandings. For one, physiological evi-dence can provide guidance as to where tolook but can not answer this question. Therehas also been a failure to understand the im-plications of a resource model: specifically, howan increase in resources could lead to an im-provement in one situation (a low initial

arousal level, unskilled subjects, etc.) but notin another (a high initial arousal level, skilledsubjects, etc.). The final misunderstandingconcerns the concepts of shared and uniquevariance. For us, the behavioral construct ofarousal is only properly applied when two ormore manipulations can be shown to sharevariance; that is, when it can be shown thatthey have similar effects on performance. Weexpect that most if not all of the arousal ma-nipulations also have unique variance com-ponents. Thus, a demonstration that sleep de-privation and impulsivity or noise and timeof day have somewhat different effects on per-formance would not cause us to reject the uni-tary arousal hypothesis. A demonstration thatthere were two clusters of variables (e.g., noiseand sleep deprivation vs. time of day and im-pulsivity) that had more in common within acluster than between clusters would lead us toreject the assumption of a single arousalsystem.

Although difficult, it would be possible totest the adequacy of larger sections of themodel as well as the overall model. To do this,we are required to specify each of the separatecovariances between observables (e.g., mea-sures of impulsivity, anxiety, and achievementmotivation, vigilance performance, reactiontime, and recall from memory) in terms ofthe hypothesized relationships between theobservables and the latent variables (e.g., effort,arousal, SIT, and STM). As long as the modelis specifiable and unique (i.e., has more esti-mates of parameters than parameters), we canobtain an overall goodness of fit of the modelto the data. Such a fit can then be comparedwith the fit of specifications of other models,some with fewer and some with more pathsto be estimated.

Before accepting or rejecting the entiremodel, it is useful to notice that the core ofour model rests on two very simple, but pow-erful, assumptions. The first is, of course, thatthe effects of the different personality traitsand of the many situational manipulations canbe interpreted in terms of only two motiva-tional states: effort and arousal. The second isthat the interrelationships of many perfor-mance tasks can be expressed in terms of theamount of SIT and the amount of STM re-sources required by the separate tasks. We be-lieve that, before the idea that there is anycommon thread to various motivational ma-


nipulations or any shared variance betweenvarious performance tasks is rejected, thepower of a two-state-two-process model shouldbe examined.

Our model may be thought of as a concep-tual factor or components analysis rotated toexperimental simple structure. That is, we be-lieve that two dimensions of motivation canbe used to summarize the interactive effectsof situational moderators and personalitytraits. Although alternative rotations of thesedimensions are possible, we believe that theuse of effort and arousal as axes minimizesthe number of nonzero relationships (paths)between the observed variables and the latentconstructs. Similarly, our use of SIT and STMas axes of information processing is a con-ceptual simple structure of the resources re-quired for cognitive tasks and, as such, is anidealized representation that emphasizes theindependence of SIT and STM components.We recognize, however, that all cognitive tasksprobably have some mixture of SIT and STMrequirements, but in varying degrees. That is,we believe that it is possible to locate the re-source requirements of vigilance, reactiontime, analogies, and memory-scanning tasksin a two-dimensional space. Furthermore, ex-perimental manipulation of the relative im-portance of SIT and STM components is pos-sible within the same task. Thus, Leon andRevelle (1983), in a study of the effects of anx-iety on analogical reasoning, varied the num-ber of elements (SIT load) and transformations(STM load) in a set of geometric analogies (seealso Mulholland, Pellegrino, & Glaser, 1980;Onken & Revelle, 1983). Similarly, the scan-ning task of Folkard et al. (1976; see also An-derson & Revelle, 1983) has alternate formsdiffering in STM load. The different forms ofthese tasks can be seen as occupying differentlocations in the resource space and should re-spond differently to motivational manipula-tions.

(1977) have identified attention, or its equiv-alent (see also Gale, 1977), and STM as pro-cesses that are differentially affected by arousal.Even more recently, Hockey (1979) and Hum-phreys et al. (1980) have discussed how thesetwo separate processes can, when combined,produce curvilinearity. Our terminology andthe analyses of dual-task situations are takenfrom Navon and Gopher (1979) and fromNorman and Bobrow (1975). Our analysis ofpersonality owes much of its inspiration to thework of Atkinson (1957, 1964, 1974) andH. J. Eysenck (1947, 1967, 1976, 1981). Therelationship between anxiety and the directionof effort has been borrowed from Mandler(1975, 1979) and Sarason (1975) and Wine(1971). Finally, it should be obvious that ourdistinction between effort and arousal bears amarked resemblance to the incentive-versus-drive distinction of Hull (1952) and of Spence(1956) and to the two motivational states dis-cussed by Broadbent (1971).17

We have not attempted to present a per-sonality theory at the level of Atkinson (1974;Atkinson & Birch, 1978) or of H. J. Eysenck(1967, 1976, 1981). As an example, we do nothave a theory of impulsivity but rather a de-scription of its correlates. Instead, we havetried to present a framework in which a moreadequate theory of achievement motivation,anxiety, and introversion-extraversion couldbe constructed. Likewise, we do not have aprocess theory of performance such as the oneproposed by Sanders (1983). Our intention,however, has been to show how personalityand motivational variables can be linked to areasonably broad range of performance tasks.As a consequence, the information-processingconstructs that we have proposed (SIT andSTM) are necessarily more general than arethose proposed by Sanders (1983). We feel thatSanders's approach complements ours and thathis conclusions support our contention that a

CONCLUDING REMARKS

We should acknowledge at this point thatmuch of what we are proposing is not original.The idea that curvilinearity can be derivedfrom opposing processes has been suggestedby Easterbrook (1959) and others. More re-cently, Folkard (1975) and Hamilton et al.

17 Part of the levels of control discussed by Broadbent(1971) can be included in thjs model if performance isseen to feedback through its effect on success and failure.Thus, effort can be seen as the higher order control processthat responds to changes in performance demands by al-locating resources. Similarly, arousal can be seen as thelower order control process that affects the availability ofprocessing resources. Thomas (1983) also discusses theeffects of feedback on subsequent effort.


limited number of motivational constructs canaccount for a substantial amount of the vari-ance in the motivation-performance litera-ture.18

What is different about our emphasis is thatwe not only try to show the interrelationshipsbetween these three personality dimensions,but our conception of a model is one in whichmultivariate and experimental psychologytruly interact. We feel that the personality lit-erature compels us to make a distinction be-tween effort and arousal. It is then up to ex-perimental psychology to find specific pro-cesses that are differentially affected by effortand arousal. Likewise, if we can discover justwhich aspect of memory is hurt by arousaland also can learn to measure this effect, thenwe would be forced to revise our ideas aboutimpulsivity and arousal if high and low im-pulsives did not differ on this measure.

In this area of personality, motivation, andperformance, we feel that it is necessary toconsider, as we have done, a broad range ofdata at one time. The literature regarding theeffects of motivation on performance is filledwith failures to replicate and even with someoccasional contradictions. If we had limitedour review to one small area in this literature,there would have been few signs of consistency.It is only when a large portion of the area,including both the personality and the exper-imental literature, is reviewed that signs ofconsistency emerge.

The advantage of our multivariate approachis that by specifying a structural model we areable to compare our theory to others (i.e., tounifactorial motivational theories or to task-specific performance theories) as well as torelate new performance tasks and personalityvariables. We are claiming that the covariancestructure of 3 personality dimensions, 6 sit-uational manipulations, and 10 performancemeasures can be understood in terms of theirrelationships to the latent variables of effort,arousal, SIT, and STM. We do not claim thatall of the variability in any single variable isaccounted for by the model, but that the modelprovides a better fit to the shared variance ofthese tasks than does a single motivational orperformance construct. Furthermore, we claimthat there is indeed shared variance betweenperformance tasks and that at least part of thisshared variance can be accounted for by the

constructs of SIT and STM. To the extent thatthis is true, we can then extend the theorybeyond the data already collected (which wereused to derive this theory) to other personalitydimensions and performance tasks. If we arepresented with a new performance measure,an analysis of that task in terms of the relativeamounts of IT and STM demands allows usto predict how that task should respond toeffort or to arousal manipulations. Similarly,by understanding how arousal and effort affectperformance, we are able to predict how othermanipulations (say alcohol or the minor tran-quilizers) should affect tasks such as sustainedmonitoring.

We now can relate personality dimensionsto situations and tasks and make specific pre-dictions about the conditions under whichpeople who differ in impulsivity, achievementmotivation, or anxiety differ in their perfor-mance in a variety of situations. We make thefollowing predictions, some of which are, aswe have already noted, empirical generaliza-tions, but most of which are only derivablefrom our model and the analyses we have pre-sented:

1. Motivationally based correlations be-tween SIT tasks can only be expected to bepositive when performance is in the resource-limited range.

2. Arousal manipulations have larger pos-itive effects on SIT tasks if the initial level of

18 Nor do we suggest why arousal should facilitate in-formation transfer or hinder short-term memory. A plau-sible model that is beyond the scope of this article couldconsider the effects of arousal as speeding up an internalclock or generally speeding up the flow of informationthrough the system. Such an increased flow of informationwould lead to improvements in IT tasks but would haveboth facilitative and debilitative effects on memory tasks.STM tasks would be hindered by the increased interferenceassociated with a higher rate of information flow. Memoryfor events learned under high arousal, however, could wellshow improvements due to what would be equivalent tomore exposure to the to-be-learned material. Such a modelcan be understood most easily by considering arousal toreduce the psychological moment. A delay of 10 s of ob-jective (external) time for a highly aroused subject wouldhave the same effect as a longer delay (say 15 s) for a lessaroused subject. Thus, in a STM task, the more arousedsubject would be expected to have less recall after a fixedinterval than would a less aroused subject. In a LTM study,however, the greater subjective time spent learning the ma-terial for the more aroused subject would more than com-pensate for the difference in recall interval.


arousal of the subjects is low or if the taskrequires more resources.

3. In the morning, high impulsives shouldexperience a greater benefit on a SIT task fromincreases in arousal than should low impul-sives.

4. High impulsives show greater improve-ments on SIT tasks with effort manipulationsthan do low impulsives.

5. After an initial delay, anxiety can im-prove performance on SIT tasks.

6. There is some aspect of STM that is di-rectly and adversely affected by arousal butnot by effort; thus, tasks that have a STMcomponent are more likely to show over-arousal deficits.

7. Low impulsives should be more likely toexperience deficits on tasks with a STM loadin the morning than are high impulsives.

8. There is a shift in the peak level of per-formance as a function of effort.

9. People who are high in achievement mo-tivation are especially susceptible to arousal-induced performance decrements.

10. High-achievement-oriented subjectscan show overmotivational deficits, but theirperformance should always be higher than theperformance of low-achievement-orientedsubjects in the same condition.

Finally, the utility of this approach shouldnot be judged solely by the accuracy of ourdecision about specific details. We hope thatthis endeavor of linking personality to moti-vational states that in turn are linked to in-formation processing accelerates the linkingof the two paradigms of scientific psychology(Cronbach, 1957, 1975; H. J. Eysenck, 1966;Vale & Vale, 1969). The model should providea common vocabulary and demonstrate tomultivariate and experimental psychologiststhat they share some common questions. Wealso see a need for them to share their para-digms. We cannot begin to understand the re-lationship between effort and arousal if welimit ourselves to univariate experiments. Evenwhen we look at two variables at once, it isfrequently not possible to reject the hypothesisthat the two variables have additive effects onsome motivational state and that motivationis curvilinearly related to performance. Astrategy that we would like to recommend isthe inclusion of a third variable. The person-ality psychologist might want to look at the

effects of two personality dimensions and ofan effort or arousal manipulation. The ex-perimental psychologist might prefer to lookregularly at one personality dimension (im-pulsivity or achievement motivation) alongwith two experimental manipulations. Ofcourse, this is no panacea, for, in addition tothe conceptual problems of just what is thecorrect linkage between these personality con-structs and motivational states, there are someserious measurement problems associated withthese personality constructs. However, such astrategy would permit multivariate and ex-perimental psychologists to make progress to-gether.

ReferencesAmelang, M., & Breit, C. (1983). Extroversion and rapid

tapping: Reactive inhibition or general cortical activationas determinants of performance differences. Personality& Individual Differences. 4, 103-105.

Anderson, K. J. (1981). The current status of the Easter-brook hypothesis. Unpublished manuscript, North-western University.

Anderson, K. J., & Revelle, W. (1982). Impulsivity, caffeine,and proofreading: A test of the Easterbrook hypothesis.Journal of Experimental Psychology: Human Perceptionand Performance, 8, 614-624.

Anderson, K. J., & Revelle, W. (1983). The interactiveeffects of caffeine, impulsivity, and task demands on avisual search task. Personality and Individual Differ-ences, 4, 127-134.

Atkinson, J. W. (1957). Motivational determinants of risk-taking behavior. Psychological Review, 64, 359-372.

Atkinson, J. W. (1964). An introduction to motivation.New York: D. Van Nostrand.

Atkinson, J. W. (1974). Strength of motivation and effi-ciency of performance. In J. W. Atkinson & J. O. Raynor(Eds.), Motivation and achievement, (pp. 117-142).Washington, DC: V. H. Winston.

Atkinson, J. W, & Birch, D. (1970). The dynamics ofaction. New \ork: Wiley.

Atkinson, J. W., & Birch, D. (1978). Introduction to mo-tivation. New York: D. Van Nostrand.

Atkinson, J. W., Bongort, K., & Price, L. H. (1977). Ex-plorations using computer simulation to comprehendthematic apperceptive measurement of motivation. Mo-tivation and Emotion, 1, 1-27.

Atkinson, J. W., & Raynor, J. O. (Eds.). (1974). Motivationand achievement. Washington, DC: V. H. Winston.

Baddeley, A. D., & Hitch, G. J. (1974). Working memory.In G. H. Bower (Ed.), The psychology of learning andmotivation: Advances in research and theory (Vol. 8, pp.47-89). New York: Academic Press.

Bavelas, J., & Lee, E. S. (1978). Effects of goal level onperformance: A trade-off of quantity and quality. Ca-nadian Journal of Psychology, 32, 219-240.

Bentler, P. M. (1980). Multivariate analysis with latentvariables: Causal modeling. Annual Review of Psychol-ogy, 31, 419-456.


Blake, M. J. F. (1967). Time of day effects on performancein a range of tasks. Psychonomic Science, 9, 349-350.

Block, J. H., & Block, J. (1979). The role of ego-controland ego-resiliency in the organization of behavior. InW. A. Collins (Ed.), Minnesota Symposium on ChildPsychology (Vol. 13). Hillsdale, NJ: Erlbaum.

Bowyer, P., Humphreys, M. S., & Revelle, W. (1983).Arousal and recognition memory: The effects of im-pulsivity, caffeine, and time on task. Personality andIndividual Differences, 4, 41-49.

Broadbent, D. E. (1971). Decision and stress. London:Academic Press.

Broadbent, D. E. (1978). The current state of noise re-search: A reply to Poulton. Psychological Bulletin, 85,1052-1067.

Browne, J. A., & Howarth, E. (1977). A comprehensivefactor analysis of personality questionnaire items: A testof twenty putative factor hypotheses. Multivariate Be-havioral Research, 12, 399-427.

Buss, D. M., & Craik, K. H. (1983). The act frequencyapproach to personality. Psychological Review, 90, 105-126.

Campbell, D. T, & Fiske, D. W. (1956). Convergent anddiscriminant validation by the multitrait-multimethodmatrix. Psychological Bulletin, 56, 81-105.

Campbell, J. B. (1983). Differential relationships of ex-traversion, impulsivity, and sociability to study habits.Journal of Research in Personality, 17, 308-314.

Cattcll, R. B. (1957). Personality and motivation: Structureand measurement. Yonkers-on-Hudson, NY: WorldBook.

Cattell, R. B. (1973). Personality and mood by question-naire. San Francisco: Jossey-Bass.

Claridge, G. (1981). Psychoticism as a dimension of per-sonality. In R. Lynn (Ed.), Dimensions of personality.Oxford, England: Pergamon Press.

Cohen, S. (1980). Aftereffects of stress on human perfor-mance and social behavior: A review of research andtheory. Psychological Bulletin 88, 82-108.

Conley, J. J. (in press). The hierarchy of consistency: Areview and model of longitudinal findings on adult in-dividual differences in intelligence, personality, and selfopinion. Personality and Individual Differences.

Craik, F. I. M., & Lockart, R. S. (1972). Levels of pro-cessing: A framework for memory research. Journal ofVerbal Learning and Verbal Behavior, 11. 671-684.

Craig, M. J., Humphreys, M. S., Rocklin, T. R., & Revelle,W. (1979). Impulsivity, neuroticism, and caffeine: Dothey have additive effects on arousal? Journal of Researchin Personality, 13, 404-419.

Cronbach, L. J. (1957). The two disciplines of scientificpsychology. American Psychologist, 12, 671-684.

Cronbach, L. J. (1975). Beyond the two disciplines ofscientific psychology. American Psychologist, 30, 116-128.

Duffy, E. (1962). Activation and behavior. New York: Wiley.Easterbrook, J. A. (1959). The effect of emotion on cue

utilization and the organization of behavior. Psycholog-ical Review, 66, 183-201.

Eysenck, H. J. (1947). Dimensions of personality. NewYork: Praeger.

Eysenck, H. J. (1966). Personality and experimental psy-chology. Bulletin of the British Psychological Society,19, 1-28.

Eysenck, H. J. (1967). The biological basis of personality.Springfield, IL: Charles C Thomas.

Eysenck, H. J. (Ed.). (1976). The measurement of per-sonality. Baltimore: University Park Press.

Eysenck, H. J. (Ed.). (1981). A model for personality. NewYork: Springer-Verlag.

Eysenck, H. J., & Eysenck, S. B. G. (1975). Manual forthe Eysenck Personality Questionnaire. San Diego, CA:Educational and Industrial Testing Service.

Eysenck, H. J., & Levey, A. (1972). Conditioning, intro-version-extraversion and the strength of the nervous sys-tem. In V. D. Nebylitsyn & J. A. Gray (Eds.), Biologicalbases of individual behavior (pp. 206-220). New "fork:Academic Press.

Eysenck, M. W. (1977). Human memory. Oxford, England:Pergamon Press.

Eysenck, M. W. (1980, November). Effects of incentiveson working memory. Paper presented at the meeting ofthe Psychonomic Society, St. Louis, MO.

Eysenck, M. W, & Folkard, S. (1980). Personality, timeof day, and caffeine: Some theoretical and conceptualproblems in Revelle et al. Journal of Experimental Psy-chology: General, 109, 32-41.

Folkard, S. (1975). Diurnal variation in logical reasoning.British Journal of Psychology, 66, 1-8.

Folkard, S., Knauth, P., Monk, T. H., & Rutenfranz, J.(1976). The effect of memory load on the circadianvariation in performance efficiency under a rapidly ro-tating shift system. Ergonomics, 19, 479-488.

Folkard, S., & Monk, T. H. (1980). Circadian rhythms inhuman memory. British Journal of Psychology, 71, 295-307.

Folkard, S., Monk, T. H., Bradbury, R., & Rosenthal, J.(1977). Time of day effects in school children's im-mediate and delayed recall of meaningful material. Brit-ish Journal of Psychology, 58, 45-50.

Frankenhaeuser, M. (1975). Experimental approaches tothe study of catecholamines and emotion. In L. Levi(Eds.), Emotions: Their parameters and measurement(pp. 209-234). New York: Raven.

Frowein, H. W. (1981). Selective effects of barbituate andamphetamines on information processing and responseexecution. Acta Psychologica, 47, 105-115.

Frowein, H. W., Reitsma, D., & Aquarius, C. (1981). Effectsof two counteracting stresses on the reaction process.In A. D. Baddeley & J. Long (Eds.), Attention and per-formance IX (pp. 575-589). Hillsdale, NJ: Erlbaum.

Gale, A. (1977). Some EEG correlates of sustained atten-tion. In R. R. Mackie (Ed.), Vigilance. New York: PlenumPress.

Geiselman, R. E., Woodward, A. E., & Beatty, J. (1982).Individual differences in verbal memory performance:A test of alternative information-processing models.Journal of Experimental Psychology: General, 111, 109-134.

Gilbert, R. M. (1976). Caffeine as a drug of abuse. InR. G. Gibbins, Y. Israel, H. Kalant, R. E. Popham, W.Schmit, & R. G. Smart (Eds.), Research advances inalcohol and drug problems (Vol. 3, pp. 47-176). NewYork: Wiley.

Gilliland, K. (1980). The interactive effect of introversion-extraversion with caffeine-induced arousal on verbalperformance. Journal of Research in Personality, 14,482-492.


Gottsdanker, R. (1975). The attaining and maintaining ofpreparation. In P. Rabbit & S. Dornic (Eds.), Attentionand performance V (pp. 33-49). London: AcademicPress.

Gray, J. A. (1972). The psychophysiological nature of in-troversion-extraversion: A modification of Eysenck'stheory. In V. D. Nebylitsyn & J. A. Gray (Eds.), Biologicalbases of individual behavior (pp. 182-205). New York:Academic Press.

Gray, J. A. (1982). Neuropsychological theory of anxiety:An investigation of the septal-hippocampal system.Cambridge, England: Cambridge University Press.

Gupta, B. S. (1977). Dextroamphetamine and measuresof intelligence. Intelligence, 1, 274-280.

Hamilton, P., Hockey, R., & Rejman, M. (1977). Theplace of the concept of activation in human informationprocessing theory: An integrative approach. In S. Dornic(Ed.), Attention and performance VI (pp. 463-486).London: Academic Press.

Hamilton, P., Wilkinson, R. X, & Edwards, R. S. (1972).A study of four days partial sleep deprivation. In W. P.Colquhoun (Ed.), Aspects of human ejfeciency: Diurnalrhythms and loss of sleep (pp. 101-113). London: EnglishUniversity Press.

Hartley, L. R. (1981). Noise does not impair by masking:A reply to Poulton's "Composite model for human per-formance in continuous noise." Psychological Review,88, 86-89.

Hebb, D. O. (1955). Drives and the C.N.S. (conceptualnervous system). Psychological Review, 62, 243-254.

Hirschberg, N. (1978). A correct treatment of traits. H.London (Ed.), Metatheories of personality (pp. 45-67).New York: Wiley.

Hockey, R. (1979). Stress and the cognitive componentsof skilled performance. In V. Hamilton & D. M. War-burton (Eds.), Human stress and cognition (pp. 141-177). London: Wiley.

Hollingworth, H. L. (1912). The influence of caffeine onmental and motor efficiency. Archives of Psychology, 22,1-166.

Howarth, E. (1976). Were Cattell's "personality sphere"factors correctly identified in the first instance? BritishJournal of Psychology, 67, 213-230.

Hughes, D. O., & Folkard, S. (1976). Adaptation to an 8-h shift in living routine by members of a socially isolatedcommunity. Nature, 264, 432-434.

Hull, C. L. (1952). A behavior system. New Haven, CT:Yale University Press.

Humphreys, M. S., Lynch, M. J., Revelle, W., & Hall,J. W. (1983). Individual differences in short-term mem-ory. In R. Dillon & R. Schmeck (Eds.), Individual dif-ferences in cognition (pp. 35-64). New \brk: AcademicPress.

Humphreys, M. S., Revelle, W, Simon, L., & Gilliland,K. (1980). Individual differences in diurnal rhythmsand multiple activation states: A reply to M. W. Eysenckand Folkard. Journal of Experimental Psychology: Gen-eral, 109, 42-48.

Kahneman, D. (1973). Attention and effort. EnglewoodCliffs, NJ: Prentice-Hall.

Kuhl, J., & Blankenship, V. (1979). The dynamic theoryof achievement motivation: From episodic to dynamicthinking. Psychological Review, 85, 239-248.

Lacey, J. L. (1967). Somatic response patterning and stress:

Some revisions of activation thory. In M. H. Appley &R. Trumbell (Eds.), Psychological stress (pp. 14-37).New York: Appleton-Century-Crofts.

Leon, M. R., & Revelle, W. (1983). The effects of anxietyon analogical reasoning: A test of three theoretical mod-els. Manuscript submitted for publication.

Lynn, R. (Ed.). (1981). Dimensions of personality: Essaysin honor ofH. J. Eysenck. Oxford, England: PergamonPress.

Loo, R. (1979). Role of primary personality factors in theperception of traffic signs and driver violations and ac-cidents. Accident analysis and prevention, 11, 125-127.

Mandler, G. (1975). Mind and emotion. New York: Wiley.Mandler, G. (1979). Thought processes, consciousness, and

stress. In Hamilton, V. & D. M. Warburton (Eds.), Hu-man stress and cognition (pp. 179-201). London: Wiley.

McClelland, D. C. (1961). The achieving society. New York:D. Van Nostrand.

Minors, D. S., & Waterhouse, J. M. (1981). Circadianrhythms and the human. Bristol, England: Wright.

Monk, T. H., Knauth, P., Folkard, S., & Rutenfranz, J.(1978). Memory based performance measures in studiesof shiftwork. Ergonomics, 21, 819-826.

Morris, L. W., & Liebert, R. M. (1970). Relationship ofcognitive and emotional components of test anxiety topsychological arousal and academic performance. Jour-nal of Consulting and Clinical Psychology, 35, 332-337.

Mulholland, T. M., Pellegrino, J. W., & Glaser, R. (1980).Components of geometric analogy solution. CognitivePsychology, 12, 252-284.

Murray, H. A. (1938). Explorations in personality. NewYork: Oxford University Press.

Naatanen, R. (1972). Time uncertainty and occurrenceuncertainty of the stimulus in a simple reaction timetask. Ada Psychologica, 36, 492-503.

Navon, D., & Gopher, D. (1979). On the economy of thehuman processing system. Psychological Review, 86,214-255.

Niemi, P., & Naatanen, R. (1981). Foreperiod and simplereaction time. Psychological Bulletin, 89, 133-162.

Norman, D. A., & Bobrow, D. G. (1975). On data-limitedand resource-limited processes. Cognitive Psychology,7, 44-64.

Norman, W. T. (1969). "1b see oursels as ithers see us!"Relations among self-perceptions, peer perceptions, andexpected peer-perceptions of personality attributes.Multivariate Behavioral Research, 4, 417-443.

Onken, J., & Revelle, W. (1983). ANATEST: A programto generate geometric analogies varying elements andtransformations. Manuscript submitted for publication.

Pachella, R. G. (1974). The interpretation of reaction timein information-processing research. In B. Kantowitz(Ed.), Human information processing: Tutorials in per-formance and cognition (pp. 41-82). Hillsdale, NJ: Erl-baum.

Posner, M. I. (1978). Chronometric explorations of mind:The third Paul M. Pitts Lectures. Hillsdale, NJ: Erlbaum.

Posner, M. I., Snyder, C. R. R., & Davidson, B. J. (1980).Attention and the dection of signals. Journal of Exper-imental Psychology: General, 109, 160-174.

Poulton, E. C. (1976). Arousing environmental stressescan improve performance, whatever people say. Aviation,Space and Environmental Medicine, 47, 1193-1204.

Poulton, E. C. (1977). Continuous intense noise masks


auditory feedback and inner speech. Psychological Bul-letin, 84, 977-1001.

Poulton, E. C. (1978). A new look at the effect of noise:A rejoinder. Psychological Bulletin, 85, 1068-1079.

Poulton, E. C. (1979). Composite model for human per-formance in continuous noise. Psychological Review,86, 361-375.

Poulton, E. C. (1981). Not so! Rejoinder to Hartley onmasking by continuous noise. Psychological Review, 80,90-92.

Rapoport, J. L., Jensvold, M., Elkins, R., Buchsbaum,M. S., Weingartner, H., Ludlow, C., Zahn, T. P., Berg,C. J., & Neims, A. H. (1981). Effects of caffeine in boysand adult males. Journal of Nervous and Mental Disease,169, 726-732.

Reuman, D. A. (1982). Ipsative behavioral variability andthe quality of thematic apperceptive measurement ofthe achievement motive. Journal of Personality and So-cial Psychology, 43, 1098-1110.

Revelle, W., Amaral, P., & Turriff, S. (1976). Introversion/extraversion, time stress, and caffeine: The effect onverbal performance. Science, 192, 149-150.

Revelle, W., Anderson, K. J., & Humphreys, M. S. (inpress). Personality and performance: Tests of the arousalmodel. In G. Mangan & T. Paisey (Eds.), Personalityand temperament: An East-West typology. Oxford, En-gland: Pergamon Press.

Revelle, W., & Humphreys, M. S. (1983). Personality, mo-tivation, and performance. Unpublished manuscript.

Revelle, W., Humphreys, M. S., Simon, L., & Gilliland,K. (1980). The interactive effect of personality, time ofday, and caffeine: A test of the arousal model. Journalof Experimental Psychology: General, 109, 1-31.

Revelle, W., & Michaels, E. J. (1976). The theory ofachievement motivation revisited: The implications ofinertial tendencies. Psychological Review, 83, 394-404.

Rocklin, T. (1981). The effects of personality, absolutedifficulty and relative difficulty on motivation and per-formance: A test of the dynamic theory of achievementmotivation (Doctoral dissertation, Northwestern Uni-versity, 1981). Dissertation Abstracts International, 42(5-B), 2138.

Rocklin, T., & Revelle, W. (1981). The measurement ofextraversion: A comparison of the Eysenck PersonalityInventory and the Eysenck Personality Questionnaire.British Journal of Social Psychology, 20, 279-284.

Sanders, A. F. (1983). Towards a model of stress and humanperformance. Acta Psychologica, 53, 61-69.

Sanders, A. F., & Reitsma, W. D. (1982). Lack of sleepand covert orienting of attention. Acta Psychologica, 52,137-145.

Sanders, A. E, Wijnes, J. L. C., & van Arkel, A. E. (1982).An additive factor analysis of the effects of sleep losson reaction processes. Acta Psychologica, 51, 41-59.

Sarason, I. G. (1975). Anxiety and self-preoccupation. InI. G. Sarason & C. D. Spielberger (Eds.), Stress andanxiety (Vol. 2, pp. 27-44). Washington, DC: Hemi-sphere.

Schalling, D. (1978). Psychopathy-related personalityvariables and the psychophysiology of socialization. InR. D. Hare & D. Schalling (Eds.), Psychopathic behavior:Approaches to research (pp. 85-106). New York: Wiley.

Seyle, H. (1976). The stress of life (Revised ed.). New York:McGraw-Hill.

Smith, B. D., Wilson, R. J., & Jones, B. E. (1983). Ex-traversion and multiple levels of caffeine-induced arousal:Effects on overhabituation and dishabituation. Psycho-physiology, 20, 29-34.

Spence, J. T, & Spence, K. W. (1966). The motivationalcomponents of manifest anxiety: Drive and drive stimuli.In C. D. Spielberger (Ed.), Anxiety and behavior (pp.291-323). New York: Academic Press.

Spence, K. W. (1956). Behavior theory and conditioning.New Haven, CTi Yale University Press.

Spence, K. W, Farber, I. E., & McFann, H. H. (1956).The relation of anxiety (drive) level to performance incompetitional and non-competitional paired-associateslearning. Journal of Experimental Psychology, 52, 296-305.

Thayer, R. E. (1967). Measurement of activation throughself-report. Psychological Reports, 20, 663-678.

Thayer, R. E. (1978). Toward a psychological theory ofmultidimensional activation (arousal). Motivation andEmotion, 2, 1-34.

Thomas, E. A. C. (1983). Notes on effort and achievement-oriented behavior. Psychological Review, 90, 1-20.

Trumbo, D. A., & Gaillard, A. W. K. (1975). Drugs, timeuncertainty, signal modality and reaction time. InP. M. A. Rabbit & S. Dornic (Eds.), Attention and per-formance F(pp. 441-454). London: Academic Press.

Vale, J. R., & Vale, C. A. (1969). Individual differencesand general laws in psychology: A reconciliation. Amer-ican Psychologist, 24, 1093-1108.

Weiner, B. (1966). Motivation and memory. PsychologicalMonographs, 50(18), 1-22.

Weiner, B. (1972). Theories of motivation. Chicago: RandMcNally.

Weiner, B. (1978). Achievement strivings. In H. London& J. E. Exner, Jr. (Eds.), Dimensions of personality (pp.1-36). New York: Wiley.

Weiner, B., & Scheider, K. (1971). Drive versus cognitivetheory: A reply to Boor and Harmon. Journal of Per-sonality and Social Psychology, 18, 258-262.

Weiss, B., & Laties, U. G. (1962). Enhancement of humanperformance by caffeine and the amphetamines. Phar-macological Review, 14, 1-36.

Wenzel, D. G., & Rutledge, C. O. (1962). Effects of cen-trally-acting drugs on human motor and psychomotorperformance. Journal of Pharmaceutical Science, 51,631-644.

Wickens, D. D., Moody, J., & Dow, R. (1981). The natureand timing of the retrieval process and of interferenceeffects. Journal of Experimental Psychology: General,110, 1-20.

Wine, J. (1971). Test anxiety and direction of attention.Psychological Bulletin, 76, 92-104.

Yerkes, R. M., & Dodson, J. D. (1908). The relation ofstrength of stimuli to rapidity of habit-information.Journal of Comparative Neurology and Psychology, 18,459-482.

Received March 23, 1983Revision received October 3, 1983

Documents

Psychological R eviewpersonality-project.org/revelle/publications/hr84.pdf · 154 M IC H A E L S . H U M P H R E Y S A N D W IL L IA M R E V E L L E m ance, w hereas at other tim