61
Acta Psychologica 53 (1983) 61-97 North-Holland Publishing Company
TOWARDS A MODEL OF STRESS AND HUMAN PERFORMANCE * A.F. SANDERS Institute for Perception Tilburg University,
TNO, The Netherlands The Netherlands
Accepted December 1982
An outline is presented of a model that aims to relate energetical and structural mechanisms of human information processing and to incorporate an interactionally defined stress notion in performance research. The first section summarizes objections to unidimensional notions of arousal and stress. It is concluded that further progress requires the analysis of joint cognitive-energetical processing structures. In the second section two current performance models - in terms of linear stages and resources - are briefly reviewed. Despite a number of problems and objections, it is decided to base the model upon linear stage notions of information processing. This is further elaborated in the third section. In accord with Pribram and McGuinness (1975) three energetical supply systems are proposed which are selectively related to specific cognitive processing mechanisms. Stress is defined as a state of imbalance in the energetical supply which is difficult to restore or to compensate. The final section is devoted to the discussion of some lines of evidence and to suggestions for future research.
1. Introduction Stress is a popular concept which is widely used in everyday life as well as i;l a variety of life sciences. However, as is often the case with such general concepts, there exist considerable divergences between the various fields in which the concept is used with regard to definition and connotation. Cox (1978) has recently described three classes of definition of stress - one in terms of a response, another in terms of a * This paper was prepared while the author held the position of Karl T. Compton, visiting professor at the Faculty of Industrial Engineering and Management, Technion, Haifa, Israel. Thanks are due to D. Gopher and A.W.K. Gaillard for valuable comments on earlier drafts, and to J.F. O’Hanlon for fruitful discussions on neurophysiology. Author’s address: A.F. Sanders, Institute for Perception, TNO, P.O. Box 23, 3769 ZG Soesterberg, The Netherlands.
000 l-69 1S/83/0000-0000/$03.00
0 1983 North-Holland
62
A. F. Sanders / Towards a model of stress and human performance
stimulus, and a final one in terms of an intervening variable - which neatly covers the set of possibilities for defining any psychological concept. It should be noted that response definitions have dominated the physiological and biochemical literature, ever since Selye ( 1936) first introduced stress to the life sciences. According to a still current physiological definition the stress concept refers to a specific response of the body to any demand made upon it (Selye 1956). If defined in this way, stress is evoked by any stimulus, the ‘stressor’, and, hence, response definitions lack the emotional component which is usually attached to the concept in common sense (e.g. Selye 1979). The emotional component is also absent in stimulus definitions, which are usually found in human performance theory. There, stress is merely a convenient label and collective noun indicating certain environmental and organismic conditions (Broadbent 197 1; Hockey 1979). Thus, a stimulus notion of stress has no pretention as an explanatory scientific concept. For instance, Broadbent could correctly speak about “the arousal theory of stress” ( 1971: 411). In the stimulus definition, noise. sleep loss and heat are all examples of “stress”. The only definition that comes close to the common sense meaning is an interactional definition, which considers stress as an intervening variable, referring to a state of unacceptable divergences between perceived demands and capabilities to adapt. A comparison between demands and achieved adaptation leads to cognitive appraisal and subsequently to a stress response if the outcome of the comparison is unfavorable and difficult to correct (Welford 1973; Lazarus 1966; Cox 1978). Attempts to cope may be concerned with changing the perceived demands, as well as.with changing the required quality of adaptation or with both. It should be noted that an exclusive emphasis on cognitive appraisal can be one-sided to the extent that it neglects conditions of direct physiological feedback about unacceptable divergences which are probably equally relevant to stress responses in the interactional sense of a reaction to failures in adaptation (see pp. 78-79). Irrespective of the level of cognitive appraisal, the interactional view emphasises the emotional component of stress. It is quite common in social and in clinical psychology and is also rapidly gaining wider acceptance. This may well be due to the explosion of absenteeism and of social benefit requests in many industrialized countries, many of which are based upon vague emotional complaints about stress in everyday life and work. Stress responses as a result of suboptimal
A. F. Sanders / Towards a model of stress and human performance
63
performance levels and demands are likely to be important contributors to this worrying development (e.g. Mackay and Cox 1979). For the theory on stress and human performance, then, there is the need for a model relating performance to stress in the interactional sense. The present paper aims to contribute to such a model. It elaborates on some earlier work on the same issue (Sanders 1980b, 198 1). First unidimensional energetical models of arousal and stress will be discussed. It will be argued that cognitive and energetical factors are so strongly intermingled that unidimensional energetical views are untenable. After a brief summary of two main performance models, an outline of a conceptual framework relating stress and performance will be presented starting from linear stage conceptions, The paper concludes with a summary of the main available evidence and with an outline of the model’s predictive potential. Readers whose main interest is in the model as such may jump immediately to section 4.
2. Unidimensional
models of stress and arousal
Traditionally stress and arousal are both energetical concepts which have very much in common. They both refer to energy mobilization of the organism and, in addition, they have usually shared the important assumptions of aspecificity and unidimensionality (Duffy 1962; Malmo 1959; Selye 1956, 1979). Arousal and stress responses are supposed to be context aspecific and to vary from little to much. This obviously raises the question how they are mutually related. In particular, if stress is defined as a response, arousal and stress are difficult to distinguish. They are almost identical because in a response definition both concepts refer to the degree of energy mobilization [l]. There only remains the traditional difference that stress is usually measured in terms of hormonal responses, while measures of arousal have been traditionally autonomous or cortical. Yet this distinction is trivial since these sets of variables are clearly interlinked (e.g. Warburton 1979). The issue of distinguishing between stress and arousal is avoided but not solved in stimulus definitions of stress; here arousal is nothing but a major explanatory concept for effects of noise, sleep loss [l] To avoid any misunderstanding the term energy should not be taken in a strict physical but rather as indicating a generalized drive state (Hebb 1955).
sense
64
A. F. Sanders / Towards a model of stress and human performance
and various other “stresses”. Only in the interactional definition do the concepts have a distinct meaning, As pointed out, stress is there a response to a perceived deficiency in performance. Consequently stress must reflect a reaction to a disturbed homeostatic state and, hence, cannot fully coincide with that state. Still starting from the assumptions of aspecificity and unidimensionality of stress and arousal, two kinds of relations have been proposed on the basis of an interactional definition. The question is whether these formulations can serve as a starting point for developing an interactional model of performance and stress. The first view starts from the Yerkes-Dodson inverted-U function relating arousal and performance. At the optimal level of arousal with regard to performance, stress is supposed to be minimal since adaptation to performance demands is maximal. At a suboptimal level of arousal, performance is also suboptimal and consequently, stress increases. The result is a U-function relating arousal and stress and an inverse monotonous relation between stress and the level of performance (Welford 1973). In contrast, the second view on the relation between stress and arousal (Cox 1978) suggests that arousal and stress are uncorrelated. Their relation depends on task demands and on intended behavior. For instance, irrespective of task demands, a low level of arousal will not be stressful to a person who intends to go to sleep. The first view appears to describe the special case where a person intends to fully cope with task demands. A family of other relations between arousal and stress can be constructed, each relation covering certain conditions of demands and intended behavior. At first sight this seems a promising approach to an interactional model of performance and stress. A major advantage is that it suggests a set of measures of stress and arousal by way of physiological and biochemical variables. Obviously a set of coherent measures would greatly enhance the theoretical status of both concepts. The late fifties and early sixties showed high expectations in this regard (Duffy 1962). Yet it should be kept in mind that the validity of the assumptions of aspecificity and unidimensionality remains a critical prerequisite to this way of theorizing. In addition the introduction of task demands as an additional variable requires a model about the relation between these demands and optimal arousal. Usually the additional property of the Yerkes-Dodson relation has been invoked that more demanding tasks have a lower optimal level of arousal. In this way qualitatively different
A. F. Sanders / Towards a model of stress and human performance
65
tasks are moulded in a single dimensional scheme of complexity or difficulty. Unfortunately, the aspecificity and unidimensionality of stress and arousal have been seriously challenged both from physiological and from behavioral research (Lacey 1967; Naatanen 1973; Hamilton et al. 1977). Although the physiological and behavioral objections are not fully independent, their distinction is quite useful. It might be possible, for instance, to have a satisfactory unidimensional arousal theory based upon behavioral data while not yet on the basis of physiological and hormonal results or vice versa. These two or perhaps three levels of analysis should not be confounded. 2. I, Physiological
objections
The physiological objections originally concerned mainly failures of finding convincing correlations between the primary psychophysiological indices of energy mobilization (Lacey 1967). It should be noted, though, that a simple notion of degree of sympathetic dominance or degree of cortical desynchronization as index of arousal cannot be expected from the interactional definition of stress. Stress and arousal will both independently affect physiological indices and, thus, low correlations would not be decisive evidence against unidimensionality of either one as long as one of them is not carefully controlled while the other is varied. This argument is additional to the usual rebuttal that low correlations are due to the fact that, besides a possible relation with arousal and stress, physiological variables also have their ‘own’ functions. Unfortunately, the physiological objections did not only concern a lack of correlation but were also related to determination of the state of the various indices by cognitive task demands. Research on event-related psychophysiological patterns has shown consistent evidence that cognitive and energetical effects are intermingled to the extent that it is preferable to speak about arousal patterns rather than arousal values, each pattern being intimately related to ongoing cognitive operations (Hamilton et al. 1977). This last conclusion implies a clear deviation from the classical sharp distinction between specific cognitive and aspecific energetical factors as determinants of behavior (Cannon 1927; Hull 1943; Hebb 1955). This distinction is not a dated conception, witness current controversies concerning cognitive models of informa-
66
A. F. Sanders /
Towards a model of stress and human performance
tion flow versus energetical capacity models (see section 3). Traditionally psychophysiological measures have been thought to purely reflect energetical state and not a mixture of cognitive and energetical elements. It has been suggested that hormonal measures more purely reflect energy mobilization since such measures might suffer less from behavioral or cognitive conditioning and therefore retain their aspecificity in comparison with autonomous and cortical measures (Selye 1979). The question is, then, to what extent catecholamines and corticosteroids, to just mention some primary candidates, reflect unidimensional arousal and stress. However, there are objections similar to those on the autonomous and cortical level. Thus, in research on stress the levels of catecholemines and corticosteroids have been found to represent different factors (Ursin et al. 1978). This may not yet surprise because the hormones have also largely noninterchangable functions. Yet, similar to psychophysiological variables, there are also suggestions concerning a relation between hormonal secretion and cognitive aspects of information processing. For instance, Mason (1975) has distinguished between a pattern 1 and a pattern 2 presumably related to the predictability of the occurrence of future events. Both patterns showed about equal elevation of noradrenaline, cortisol, thyrotropine and thyroxine but differed with respect to adrenaline secretion. Adrenaline was not elevated in the case of high predictability and may represent therefore a response to an unexpected event. Mason (1975) suggests the description of patterns of stress responses rather than measuring degrees of stress as the main task for future research. This suggestion is much along the lines proposed for arousal and reminds one of Appley and Trumbull’s (1967) question whether the stress concept has done anything but replacing older notions like anxiety, frustration and conflict. 2.2. Behavioral
objections
As argued above, the physiological objections, serious as they are, should not be used against unidimensional notions of arousal and stress as explanatory concepts for behavioral results. In view of the current stimulus definition of stress in human performance theory the discussion is necessarily restricted to evidence on the unidimensional arousal concept, the inverted-U relation between arousal and performance and the effect of task complexity on optimal arousal. This has been the
A. F. Sanders / Towards a model
of stressand
human performance
61
traditional model for explaining effects of abnormal conditions such as noise, sleep loss, heat, strong incentives and psychotropic drugs. There is no need repeating Broadbent’s (197 1) summary of the support as well as the problems for the model from studies on the interactive relations between the effects of sleep loss, noise and incentives on performance. They fit the unidimensional theory in that noise and incentives enhance and sleep loss decreases arousal. Thus a negative effect of noise would be due to overarousal and of sleep loss to underarousal. Noise and incentives have been found to have positive effects on performance after sleep loss, while a combination of incentives and noise has a negative effect after normal sleep. Yet, Broadbent (197 1) has also noted problems for a simple unidimensional model. One concerns the fact that noise has only detrimental effects on performance during the later part of a work period, while arousal theory predicts a stronger effect of noise in the beginning of work. The case in point is that arousal is supposed to decline as a function of time-on-task. Hence, if noise is ‘overarousing’, its strongest effect should occur in the beginning of a work period. Another and perhaps very serious problem is that the pattern of interactions between effects of stresses is by no means a general phenomenon. It has been found in self-paced serial reaction tests and, to some extent, in vigilance tests but there are other tasks which do hardly suffer from either sleep loss or noise. The inverted-U hypothesis could handle some of this by stating that more complex tests have a lower optimal level of arousal and that ‘interesting’ tests are intrinsically arousing. Yet this is at best an emergency explanation (Wilkinson 1969), which in critical tests can be shown to be bluntly wrong (Sanders et al. 1982). Apparently the type of task demand is a highly relevant determinant of the results. This last conclusion has also been strongly underlined by research suggesting changes in performance as expressions of shifts in strategy rather than of quantity of output. What kinds of performance changes may occur, depends on the type of the required cognitive operations as well as on the pattern of activation. As argued by Hamilton et al. (1977), the organism should be conceived of as being in one of several arousal states which are intimately related to ongoing cognitive operations. For a theory of stress and performance this means that the analysis cannot be limited to simple energetical dimensions. The cognitive demands should be considered as well, despite the fact that this means an enormous complication.
68
A. F. Sanders / Towards a model of stress and human performance
3. Performance models This last conclusion requires consideration of conceptual models of human information processing in order to clarify the nature of the ongoing cognitive operations and their accompanying states of arousal. For the present aim the analysis will be limited to an outline and some rough features of two major conceptual frameworks of performance, i.e. resource allocation models and linear stage models. This choice and in fact the whole discussion of this section is stimulated by Rabbitt’s (1979) analysis of suitable current performance models for research on stress. The distinction between’the frameworks is by no means new. It can be traced back to Ktilpe (1895) and to Donders (1868), but has been dormant for a considerable time. Both types of models are often conceived of as competitors in the analysis of human performance in general (Logan 1978), and in particular with regard to assessing effects of stress on performance (Rabbitt 1979). However, it will be argued that the models have a different emphasis and are partly concerned with different questions (see Gopher and Sanders 1983, for a more detailed discussion). Generally speaking, linear stage models are characterized by an interest in building a systems architecture. Therefore their emphasis is mainly on cognitive computational processing mechanisms. In contrast, capacity models put most emphasis on strategical allocation of attentional and, hence, energetical resources to various mental functions. In a way the models reflect the sharp distinction between cognitive and energetical factors which was touched upon earlier. Yet both have wider claims. 3.1. Linear stage models The basic aim of these models is a description of information flow through the organism as a sequence of processing stages mediating the transformation from signals into responses. It is the implicit model underlying the many box and building block conceptions. Donders’ (1868) original and still current idea (e.g. Pachella 1974; Salthouse 1981) was to obtain estimates of the time taken by a processing stage through subtraction of RT to a task that contains n stages and another task that contains n - 1 stages. As such the proposal failed because it proved extremely hard to find tests which differed with respect to one
A. F. Sanders / Towards a model of stress and human performance
69
and only one stage, and also because it is quite doubtful whether the stages remain really identical. There are several remaining “identical” examples in the literature which show that these are often untenable assumptions (Gopher and Sanders 1983). Despite these problems the linear stage conceptual framework has retained considerable appeal (e.g. Welford 1967; Smith 1968). Yet its surge during the seventies was undoubtedly due to the formulation of the additive factor logic (Sternberg 1969). In contrast to the subtraction method where stages were postulated a priori, the additive factor method attempted to infer stages by determining the mutual relations between the effects of task variables on RT. Additive effects of variables will usually mean that separate stages are affected, while interactions usually mean that the two variables affect at least one common stage. The method is not watertight and, therefore, deductions from actual experiments should be carefully phrased (Sanders 1980). The problem is that the additive factor method is based upon a number of strong assumptions including unidimensional cognitive processing, strict serial processing between stages, no feedback loops during the reaction process and a constant stage output. As soon as some of these assumptions are weakened the data. become multi-interpretable (Taylor 1976; McClelland 1979). In fact it is simple to show that none of the assumptions is generally valid, even not within the context of choice reaction processes, the exclusive additive factors paradigm. Several examples are discussed by Gopher and Sanders (1983). Yet this may not be immediately decisive. Instead, it is relevant to determine the limiting conditions where the assumptions of the logic apply (Posner 1978). It is important that there are inbuilt criteria for deciding whether this is the case in a specific experiment. The next step, then, is to find out to what extent the architecture of the system, derived from results within the limits of the method is useful for and valid to conditions beyond its scope. These issues will be further discussed in section 5.6. For the present it suffices to note that, within the area of traditional choice reaction processes, the research has shown a consistent picture of processing stages (Sanders 1980). Thus in the choice reaction area extension to research on the effects of traditional state variables seems appropriate. The simple rationale is that processing stages are supposed to reflect different sets of computational process in the brain. Hence they might as well qualify as typical
70
A. F. Sanders / Towards a model of stress and human performance
“ongoing cognitive operations” to which different states of arousal are related. This last hypothesis expects that effects of suboptimal environmental or organismic conditions selectively interact with or add to the effects of the computational variables. There could emerge relevant cognitive-energetical pictures if stable relationship between computational and state related variables show up. Even beyond the interpretative framework of the additive factor logic such patterns could provide useful empirical information. Relations between task variables are investigated and that means a level of abstraction beyond the common practice of investigating state effects on individual tasks (Sanders and Bunt 1971). 3.2. Capacity models The Wtirzburgers (e.g. Kiilpe 1895) challenged Donders’ stage approach by arguing that variations in task can never be described in terms of simple stage deletion or addition. Instead changes in the task are likely to cause qualitative changes in the total of processing. In somewhat different wording this is today still the position of capacity or resource theorists of human performance. As a broad statement adherents of the additive factor logic would also sympathize with this view since two tasks might well have different stage structures and, therefore, be qualitatively different. The point is obviously what happens when the difference between the tasks is limited to the value of one experimental variable. According to linear stage models this difference should not affect the stage structure but only the processing duration of a subset of the stages. According to resource models this still might cause shifts in strategical control of the task and, consequently, in resource allocation to the various operations concerned. Yet it is relevant to note that capacity theories differ among themselves with regard to changes in either resource volume or in resource strategy as a consequence of stresses. A resource volume theory tends to speak in terms of “more or less” resources available for a certain task with actual levels of task output as criterion (e.g. Navon and Gopher 1979). As less resources are available, task output is less but not Thus an important assumption of resource qualitatively different. volume theory is process invariance: the nature of the processes is not affected by the allocated resource volume. Only the total output may change. The relevance of this axiom is clear when it is realized that
A. F. Sanders / Towardr a model of stress and human performance
71
resource volume theory has a main interest in the analysis of dual task performance with relative priority as a variable. This type of analysis is impossible if the nature of the tasks changes as more or less resources are allocated to one of the tasks. In contrast, a resource strategy model does not assume invariance of the nature of the operations. Instead the operations may undergo fundamental changes as a function of practice (Bainbridge 1978) processing priority or information load (e.g. Sperandio 1972; Rabbitt 1979). With regard to stress and arousal strategical models emphasize changes in loci of control and qualitative shifts in performance (e.g. Monk and Leng 1982). Hence a strategical theory is not primarily interested in spotting quantitative effects of stresses on output, but foremost in qualitative changes of strategical control like the mental block (Bills 1931), neglect of peripheral elements (Bartlett 1953), emphasis on handling recent or prerecent items (Hamilton et al. 1972) and levelling or sharpening of allocation priorities (Hockey 1979). Qualitative changes in the processes as a function of strategy is actually a basic element of strategical resource theories. Both versions of resource theory distinguish sharply between automatic‘and controlled processes; the first type is passive and requires at best minimal resources in order to operate; the second type is active and requires ample resources (Norman and Bobrow 1975). In particular the strategical versions of capacity theory suggest that a linear stage approach to stress is doomed to fail since (1) it is limited to a single quantitative time measure; (2) it is predominantly concerned with ‘data-driven’ models; (3) it starts from untenable assumptions; and (4) the majority of ecologically relevant tasks are beyond its scope (Rabbitt 1979). Although in some experiments subjects might be forced to operate within the constraints of the additive factor assumptions the effects of environmental stresses. observed there, would qualitatively differ from those obtained under conditions of free strategical control of resource allocation. Moreover, this latter type of condition is a better approximation of reality. Some of these objections are undoubtedly valid and have been already raised in the paragraph on the subtraction method. A more detailed discussion is found in Gopher and Sanders (1983). For the present the question is whether the objections are sufficiently serious to apriori forego the exercise of applying linear stage thinking to performance and stress. In defense of linear stages three issues may be raised.
12
A. F. Sanders / Towards a model of stress and human performance
First, it is debatable whether in all fairness linear stage models can be viewed as merely passive and hence data-driven information processors. In the analysis of stages in choice reaction processes (Sanders 1980) there are various examples which all concern active attentional and self-initiated processes. A detailed discussion will be postponed to the next section; yet one example concerns presetting response choice on the basis of expectancy; another example is preparatory motor adjustment if the moment of arrival of the signal can be accurately predicted. On the other hand it is admitted that these active processes operate within the constraints of the stage structure and the assumptions of the theory. As soon as these are no longer valid the additive factor logic fails. It is well understood, therefore, that aspects of strategical control are at best partly covered by this method; the logic fails as soon as performance changes are not exclusively expressed in processing duration. Second, as was already briefly mentioned, the additive factor logic has a built-in check regarding the adequacy of its application. If under ‘normal’ conditions of testing two variables show consistently additive effects on RT while an interaction appears when the test is carried out under some abnormal condition, it should be seriously considered whether the abnormal condition has induced a change in stage output. For instance, the abnormal condition might have led to incomplete processing during a stage or to the operation of internal feedback loops between the stages. In fact, such changes would be operationally identical to qualitative changes in locus of control, as suggested by Rabbitt (1979) and Hockey (1979) and, if found to occur generally. they would seriously undermine the application of linear stage models. The point is that stage models require a fair degree of stage robustness, i.e. invariance of the stage structure when complicating conditions are introduced. Third, it should be noted that linear stage and resource volume models have a fair degree of communality. If resources refer to attentional energy supply to computational structures, which are the devices which actually consume the energy, the two types of framework easily converge. Allocation of capacity, then, means the extent to which certain computational structures are selectively activated and ready for processing. Processing time is determined by energetical resource allocation as well as by the remaining computational “work” required by the operations contained in a stage (Gopher and Sanders 1983).
A.F. Sanders / Towards a model of stress and human performance
13
One of the constraints imposed by linear stage theory is that in single task performance, capacity may not be freely distributed over stages. A free distribution would mean that by emphasizing one stage a deficit in resources may arise at the next stage (Moray 1967). The point is that this invariably leads to interactions, and, again, if that would be the general phenomenon, the linear stage model fails. It should be noted, though, that free distribution of resources from a single pool over the operations contained in one task would also affect the requirement of process invariance, as assumed by resource volume theory. Hence multiple resource volume and linear stage models converge considerably easier than do single resource volume and linear stages. The present trend in resource volume theories tends to assume multiple resources (Navon and Gopher 1979; Gopher and Sanders 1983); multiple resources come close to the idea of energetical mechanisms, serving exclusively some but not other types of mental operations. Despite the convergences between resource volume and linear stage models, there also remain pronounced differences in assumptions, methodology and primary centers of interests. The resource volume theory is foremost interested in phenomena of perceptual-motor load, in time sharing performance, in automatic versus controlled processing, and in strategical principles in dividing attention between tasks. In contrast, linear stage models are directed to more limited questions about the systems architecture in the information flow within one task and, then, primarily limited to choice RT tasks. The argument is dealt with in greater detail by Gopher and Sanders (1983), who arrive at the conclusion that the frameworks are complementary rather than competitive. The situation is very different with regard to resource strategy notions; there the term resource is defined very widely and becomes nothing but a vague reference to almost any processing capability, energetical as well as structural. This is clear from formulations like resources are “such things as processing effort, the various forms of memory capacity and communication channels” (Norman and Bobrow 1975: 45), or, again, resources are “acquired information about the structure of particular tasks and about the external world which are used by the subject in order to actively control their momentary perceptual selectivity and their choice of responses” (Rabbitt 1979: 129). The problem is that nothing is specified in such a formulation and
74
A. F. Sanders / Towarak a mode/ of stress and human performance
that any result can be considered as support. A full emphasis on strategical changes and control without assumptions about structural constraints carries the danger of making formal theory building completely impossible (Gopher and Sanders 1983). Obviously, resource allocation strategy models do not drive at that extreme, witness the formulation of principles of strategical change on the basis of stress research (Hockey 1979). It will be an interesting exercise to relate these principles to the results derived from experiments on linear stage models.
4. A cognitive-energetical
model of arousal, stress and performance
The apology of the previous section was meant to collect sufficient courage to embark on an outline concerning a model relating energetics and processing information, starting from a linear stage conceptual framework. As argued the starting point is the assumption that the duration of processing in a stage is affected by the state of the subject as well as by computational demands. Sanders (198 1) has briefly described an outline of the model. It relies upon four computational stages in the traditional choice reaction process, that appear best established by way of additive factor studies. These stages are stimulus preprocessing (affected by signal intensity), feature extraction (affected by signal quality), response choice (affected by S-R compatibility) and motor adjustment (affected by time uncertainty). As usual, the names attached to the stages are complimentary, but the accompanying variables have all shown to have mutually additive effects in at least two different studies each (Sanders 1980). A relevant assumption of the model is that there are only effects of the subject’s state on processing duration to the extent that active processes play a role in the cognitive operations of a stage. This assumption is in line with the current notion that active or controlled processes are energy or resource dependent while passive or automatic processes are not (e.g. LaBerge 1975; Logan 1979). 4.1.
Three types of energetical supply
In this section it will be argued that the computational stages, as briefly outlined above, rely upon three types of energetical supply or resources.
A. F. Sanders / Towards a model of stress and human performance
15
In line with the notion of multiple resources, the processes involved in different stages draw upon different energetical resources. A first resource type is related to motor adjustment, a second to feature extraction and a third one to response choice. Stimulus preprocessing finally, is only dependent upon automatic processes and therefore does not require a separate energetical resource. This is not to say that stimulus intensity is irrelevant to energetical state. On the contrary, it is traditionally considered as arousing and “attracting capacity” (Kahneman 1973). In its role of collative variable (Berlyne 1969), it has been further considered as essential to energizing the organism (Hebb 1955). With regard to motor adjustment, the efficiency can be optimized by preparatory processes and timing, which can be assumed to “preset” motor adjustment as close as possible to the “motor action limit” (Naatanen and Merisalo 1977). There is now considerable evidence that preparation is time consuming and voluntary (e.g. Sanders 1972) can be maintained for only a short period of time (Gottsdanker 1975) and is difficult to combine with other processes (e.g. Sanders 1971; Posner and Boies 1971). In fact, preparatory processes leading to optimal presetting are close to what Posner (1978) has called alertness as a component of attention. Posner has discussed the effect of time uncertainty in terms of ‘receptivity to external signals’, but there are strong arguments favoring an interpretation by way of motor preparation, albeit on a central level (Gaillard 1978; Sanders 1979). The effect can be considered to reflect a general phasic motor readiness to respond, which is confirmed by physiological evidence about brain potentials and behavioral evidence about interaction between the effects of time uncertainty and instructed muscle tension on reaction time. In addition preparatory processes probably also depend on processes of tonic alertness (Posner 1978) in view of observed interactions between effects of time uncertainty, sleep state, amphetamine administration and timeof-the-day on reaction time (Frowein 1981a). Signal quality is supposed to affect processes of feature extraction as required for identifying the signal. Presumably feature extraction and encoding require energetical supply that differs from that needed in motor adjustment. At first sight it may be doubted whether encoding processes require any resources at all in view of the ample evidence that they are predominantly ‘bottom up’ and run off automatically. Thus they do not always appear to profit from focused as opposed to divided attention (Egeth et al. 1972; Shiffrin 1975), they do not appear to affect
16
A. F. Sanders / Towards a model of stress and human performance
probe RT to a concurrent signal (Posner 1978), they are difficult to filter out (Logan 1980) and there are conditions where ‘automatic pathway activation’ leads to benefits but not to costs (Posner and Snyder 1975). However, none of these findings appears to hold in the case of unfamiliar or degraded signals (Comstock 1973; Proctor 1978) and, in addition, there is also accumulating evidence that, despite automatic processing, encoding can profit from attentional orienting (Posner 1980; Sanders and Reitsma 1982a). In short, there is ample evidence that feature extraction can profit from energetical resources. The active processes, involved in encoding, may have the primary role of separating relevant from irrelevant elements of percepts. This may be almost not needed in the case of familiar signals but very important when a global analysis must be complemented by a local one to produce a satisfactory stage output (Navon 1977; Broadbent and Broadbent 1980). To the extent that active feature extraction is involved, these active processes refer clearly to Posner and Boies’ ( 1971) notion of selective attention. A third type of energetical supply is probably required for handling incompatible S-R relations or, more generally stated, for adequate functioning of the response choice stage. This stage is the link between perception and action and comes closest to handling decision rules, reasoning and related cognitive processes which are subsumed by Posner (1978) as conscious processing and constitute his third component of attention. It is probably of interest to the present argument that the later a stage is in the suggested sequence, the more its active processes refer to “presetting the stage” prior to the presentation of the stimulus. This is very clear for motor preparation, still quite evident for response choice but much less well established and probably much weaker in feature extraction. Whatever the difference between these types of activity, this discussion fits the earlier argument that there is ample space for active processes in linear stage models, contrary to Rabbitt’s (1979) suggestion that such models are merely ‘data-driven’. There is evidence for stimulus-independent activity enabling anticipation and readiness to respond; for activity in signal analysis enabling perceptual cleaning and selection upon arrival of signals and, finally, there is evidence for resources relating to control of response choice and handling decision rules.
A. F. Sanders / Towardr a model of stress and human performance
4.2. Resemblance
-II
to other notions
Throughout this section the resemblance between the types of resources and Posner’s components of attention has been emphasized. In addition the resemblance to Welford’s (1973) three criteria for optimal energetical conditions may be noticed: optimal stimulation (stimulus analysis), optimal patterning of events (readiness to respond) and optimal conflict (decision load). Yet, the most striking resemblance concerns probably the mechanisms as discussed by Pribram and McGuinness (1975) on the basis of neurophysiological results obtained through clinical and animal studies. Pribram and McGuinness consider three systems in the control of attention, namely an arousal system as a phasic response to input, an activation system as a tonic readiness to respond and, finally, an effort mechanism as a coordinating and organizing principle. Effort is supposed to coordinate the activity of arousal and activation, but has in addition the wider function of promoting the competence of the information processing system and, as such, it comes close to the driving force behind reasoning and decision making. Pribram and McGuinness sought to define the major neuroanatomical structures involved in sustaining activity within each of these three systems, as well as interconnections within and between them. Arousal was said to reside within the classic core portion of the neuroaxis comprising the mesencephalic reticular formation with ascending and descending projections, sensory projection areas of the cerebral cortex and connected nuclei within the non-specific (intrinsic) cortex, thalamus and hypothalamus. The amygdala was assigned a particularly critical coordinating function between the cortical and subcortical elements of the arousal system. Activation mainly involved the brain’s motor control and coordinating structures with integration of this function assigned to the basal ganglia, particularly the corpus striatum. Finally, effort was said to be a function of the limbic system in a circuit involving the cingulate cortex, hippocampus, septal nuclei, posterior hypothalamus and the anterior thalamic nucleus. The main bridge between arousal and effort was from the sensory cortex and amygdala to the hippocampus; and between effort and activation, from the cingulate to the corpus striatum. Their scheme is reminiscent of earlier neuroanatomical models proposed by Routtenberg (1968) and Gray (1971). The latter made no distinction between what Pribram and McGuinness called arousal and
78
A. F. Sanders / Towards a model of stress and human performance
activation. Both processes were said to reside within the cortex under the influence of two subcortical systems. The first was the classic Ascending Reticular Activating System (ARAS), and the second a system comprising limbic structures. The ARAS was seen to respond to extroceptive stimulation in a manner directly influencing cortical activation. The limbic activating function was said to be reciprocally linked to that of the ARAS, under normal waking circumstances, so that excessive or deficient extroceptive stimulation would not lead to variable cortical activation. Thus the activating function ascribed by the earlier theorists to the limbic system differs to no large extent from effort as defined by Pribram and McGuinness. 4.3. A model of stress and arousal Coupling these neurophysiological notions to those derived from the linear stage model delivers the main elements of fig. 1. It may be a little premature to assign the effort mechanism the double function of controlling and coordinating arousal and activation as well as directly controlling ‘conscious’ processing (Posner 1978). This last function is indicated by the direct connection between effort and response choice (see also Gaillard 1980) as well as between effort and evaluation. Together this refers to the earlier discussed role of effort in reasoning and decision making. Apart from directly modulating arousal and activation, effort is also capable of uncoupling the direct connection between these mechanisms, as suggested by Pribram and McGuinness (1975) [2]. Fig. 1 has another feature, which has not yet been properly touched upon. In order to be informed about the state of arousal and activation, there must be a mechanism evaluating their appropriate functioning. There may be at least two types of information which the evaluation mechanism receives. One concerns direct feedback reflecting the physio-
(21 An energetical mechanism with a main function of controlling and coordinating other energetical mechanisms is of course in no way new. In fact it fits completely classical notions on maintaining equilibria (e.g. Cannon 1927) and in the more recent psychological literature it has been proposed by Broadbent (1971) as an upper-arousal mechanism. In fact it has been a recurrent theme (e.g. Wilkinson 1963; Gaillard and Trumbo 1976; Hockey 1979: 165) but, at least to my knowledge, it has not been put in proper theoretical perspective. For instance in the first major theory on effort (Kahneman 1973), effort and arousal were supposed to covary in the sense that the volume of effort depended on the state of arousal.
A. F. Sanders / Towar& a model of stress and human performake
79
evaluation mechanism
energetical mechanisms
processing stages experimental variables
stimuli mtenslty
signal quality
S-R compatibility
time uncertainty
Fig. I. A cognitive-energetical linear stage model of human information processing and stress. The cognitive level consists of computational processing stages derived my means of the additive factor method (Stemberg 1969). There are three energetical supply mechanisms, two of which are basal (arousal and activation) and coupled to respectively input and output processing stages. The basal mechanisms are coordinated and supervised by effort, which is also directly linked to the central stage of response choice. Apart from direct energetical supply to this stage, effort serves the function of keeping the basal mechanisms at an optimal value. Information about the state of the basal mechanisms is mediated by an evaluation mechanism. For further details, see text.
logical state of the system, so as to guarantee immediate action of effort in case of imbalance. Another type of feedback reflects the level of performance on a cognitive level. The criteria of the evaluation mechanism then, lead to an evaluation of the adequacy of performance as well as of the state of organism. There are probably multiple criteria containing short-term instructions, long-term aims and goals, as well as physiological data about the range of variation the bodily mechanisms can bear. This type of evaluation mechanism was a major element in Kahneman’s (1973) theory of effort. This brings us close to delineating a cognitive concept of stress in human performance theory: Stress will arise whenever the effort mechanism is either seriously overloaded over time or falls altogether short in accomplishing the necessary energetical adjustments. In this formulation, stress and effort covary to the extent that continuing high demands on effort without sufficient success in maintaining or restoring an equilibrium are supposed to constitute the basis of stress responses. Thus allocating effort does not evoke stress per se. It is the presence or the threat of a lasting disturbance of the equilibrium which is essential.
80
A. F. Sanders / Towards a model of stress and human performance
Fig. 1 suggests various loci of stress: it may arise because effort fails in correcting the effects of too high or too low a level of arousal, too high or too low a level of activation, while, finally, there may be failures to supply sufficient energetical resources to reasoning and decision making. Altogether this would imply at least five patterns of stress. The converging element of these patterns is a deviant state of the evaluation system, which may cause a common subjective feeling of ‘stress’. 4.4. Five patterns of stress The case of overstimulation of the arousal system may reflect Selye’s (1956) shock phase of the general adaptation syndrome. Effort is needed to redress the arousal mechanism in order to protect the organism from the effects of overstimulation. In everyday life this may include states of panic, arising from sudden unexpected and threatening stimulation as well as less dramatic cases of strong stimulation. A major consequence of overarousal could be a direct energetical overflow to the activation system which in turn promotes immediate action without well-considered cognitive analysis (see fig. l), Hence an immediate uncoupling between arousal and activation is required in order to prevent a vicious circle. Ethics usually prohibit human studies on this topic but, in a much reduced form, the phenomena of defensive reactions to loud stimulation (e.g. Lynn 1966; see section 5) do probably bear on the issue. The opposite case of understimulation of the arousal system is presumably concerned with phenomena of habituation as commonly found in repetitive and boring tasks (e.g. O’Hanlon 1981). Decrements in perceptual intake of information during vigilance tests are a common example (e.g. Mackworth 1969). It should be noted that research during the sixties has emphasized the difference between effects of perceptual habituation and of reduced readiness to respond in vigilance tests along the lines of d’ and /3 in signal detection theory (Broadbent 1971). As long as under-arousal is considered, the effects on d’ are implied, while effects on p refer to under-activation. In everyday practice they are probably subsumed under the common heading ‘boredom’; yet it is relevant to distinguish two sources, despite the fact that effects on d’ and p may not invariably refer to respectively perceptual and response factors (Sanders 1977: 6). The case of overactivation is probably identical to a hyperactive
A. F. Sanders / Towards a model of stress and human performance
81
nervous state as described by arousal theorists in the fifties. In everyday life it may be especially related to anticipatory stress as commonly described to occur when anticipating threatening situations. A final state of stress may be related to direct failures of decision making and reasoning. It concerns intolerable ambiguities in deciding or problem solving. Such ambiguity is supposed to lead to sustained high effort and, if the problems are not solved, to the common conflict type of stress. It remains of course to be seen to what extent these patterns make sense in terms of experimental evidence rather than intuitive description. One relevant issue in this respect concerns the assumption of relative independence of arousal and activation as energetical supply systems as opposed to the assumptions about their mutual interplay. To what extent will it be possible to separately investigate the functioning of arousal, activation and effort since the total of their effects is usually observed? In particular, will it be possible to decide whether performance effects are due to deviant energetical supply stemming directly from arousal or activation or, indirectly, from deficient correction through effort? In short, what are the lines of evidence and what future research programs are envisaged? Most of the present evidence bearing on the model stems from research of effects of suboptimal conditions on performance. This research is more relevant to the general energetical features of the model than to the deductions about stress. On this last topic only plans and suggestions for research can be outlined.
5. Some lines of evidence The most general prediction of the model is that suboptimal conditions have specific rather than general effects on reaction processes. Without evidence for selective relations between “state” variables and the proposed energetical mechanisms the model does not make sense as a framework. In turn the energetical mechanisms are supposed to have direct ties with specific computational mechanisms. Hence the research concerns simply the relations between the effects of “computational” and “state” variables on reaction processes. An interaction between a state and a computation related variable is interpreted as an effect of that state variable on the energetical supply required for the involved
82
A. F. Sanders / Towards CImodel of stress and human performance
processing stage. In other words as an effect on arousal, activation or effort. The studies, which are directly relevant, are on effects of sleep loss, amphetamines and barbiturates and their relations to effects of computational variables affecting choice reaction time. Indeed, the results suggest selective rather than general relations between these classes of variables. In the following they are briefly reviewed. 5.1. Effects of suboptimal conditions
First, then, Frowein (1981a, 1981b, 1981~) observed that the effect of amphetamine interacts with that of time uncertainty and with some variables affecting the speed of execution of the motor response. In addition he found that the effect of amphetamine was additive to that of various major input variables like signal intensity and signal quality. Together this suggests a selective effect on the output side of processing. Frowein found also weak evidence for an interaction between the effects of amphetamine and S-R compatibility. Yet, while the stimulant drug had the effect of speeding up motor adjustment and movement time, the time taken by response selection tended to slow down. A final finding on amphetamines was that this drug had its most convincing effect when performance was suboptimal as a consequence of, say. prolonged work or sleep loss; amphetamine appeared to counteract a decrement of performance rather than raising performance under optimal conditions, this with the exception of the motor effects. The effect of a barbiturate on choice reaction processes was quite different. In a long series of studies Frowein observed that the effect of a barbiturate was additive to that of all his computational variables except for that of signal quality. The usual effect of the barbiturate was to slow down response time and this was more pronounced when reacting to a degraded than to an undegraded signal. This result was further substantiated by a selective effect of the barbiturate on the Nz component of the evoked potential (Frowein et al. 1981). Together, these results suggest a selective effect on the input side. The effects of sleep state, finally, proved to have two major interactions with those of computational variables namely signal quality and time uncertainty (Frowein 1981 b; Sanders et al. 1982). In addition, additive effects of sleep state were found to those of S-R compatibility and signal in-
A. F. Sanders / Towardr a model of stress and human performance
83
tensity. Converging evidence for these results was obtained by Sanders and Reitsma (1982a, 1982b). In summary, it can be said that there is evidence for selective rather than general effects of suboptimal conditions on choice reaction time, which is consistent with the essential assumption of the model. The data find the most simple interpretation by assuming an effect of barbiturate on arousal, of amphetamine on activation and of sleep state on both (fig. 2) [3]. 5.2. Loci of effects Fig. 2 suggests that the discussed suboptimal conditions have an immediate effect on arousal and activation and, through there, on processing duration of corresponding stages. However, the question may be legitimately raised whether the effects cannot be ascribed directly to computational processes. Still another alternative is that the effects are on effort rather than on arousal or on activation. How to distinguish between these three alternatives? The iirst question is how direct effects on computational stages can be distinguished from indirect effects which depend on variations in energetical supply. One important way of deciding may be that direct computational factors usually have their effect on all individual trials and, hence, shift the whole distribution of reaction times. Energetical factors, on the other hand, have the characteristic that their effect may strongly vary between individual trials. Therefore their effect is usually most pronounced at the higher end of the distribution and may indeed be absent at the lower end (e.g. Sanders and Hoogenboom 1970). Obviously, this requires an analysis of higher moments than mean reaction time. Another way of deciding is that in the case of direct computational effects one would not expect that they strongly depend on motivational variables such as time-on-task, knowledge of results, time pressure or other means of affecting the evaluation mechanism. With regard to effects of drugs and sleep loss there is sufficient evidence favoring an energetical interpretation. This effect is usually
[3] This interpretation is obviously limited to the small doses of amphetamine and barbiturate, as used by Frowein. For example it is widely known from pharmacological research that large doses of barbiturate have a quite general action on the central nervous system. The present results suggest that dose and degree of selectivity of the barbiturate effect are related.
84
A. F. Sanders / Towards a model of stress and human performonce
=
I
effort
=
1 activation
arousal
1,
II
1 feature extraction
f barbiturates Fig. 2. An interpretation stages in choice reaction
motor adjustment
----_------_
7
sleep-state
f
of the effects of barbiturate, tasks.
7
amphetamtnes amphetamine
and sleep state on processing
much more pronounced at the high end and quite susceptible to motivation. The second distinction concerns effects on either arousal or activation versus effects on effort. Both are feasible: for instance, sleep loss may cause malfunctioning of arousal and activation - this was in fact the first interpretation of fig. 2 - and if this malfunctioning is not corrected by effort, there is a decrement of performance. Alternatively arousal and activation might fall short in supplying optimal energy in the experimental as well as the control condition of a study on sleep state. Suppose that only after normal sleep subjects are capable of correcting arousal and activation to their optimal value by way of effort. In that case sleep loss would also have an adverse effect on performance but this time the effect of sleep loss would be due to inadequate correction of arousal and activation by way of effort. It will be argued that a principal way of deciding between these alternatives is again by considering effects of motivational variables like time-on-task and knowledge of results. The main argument is that such variables are strongly related to effort allocation and, hence, their effects provide important tools for distinguishing between effects on the more “basal” mechanisms of arousal and activation and the more voluntary mechanism of effort. Additional criteria are that an effect of a suboptimal condition on effort would (1) equally affect energetical supply correction of arousal and activation, and (2) affect response choice variables. Both criteria are not watertight, however: the relation between effort and response choice is among the least established
A.F. Sanders / Towards a model of stress and human performance
85
elements of fig. 1 and it is not necessary that both arousal and activation always need correction at the same time. Yet specific relations of a suboptimal condition with one of these energetical mechanisms is certainly indicative. 5.3. Effects of time-on-task (TOT) and knowledge of results (KR) The effect of TOT on reaction performance - detection as well as time _ is properly summarized by Mackworth’s (1964) square root relation. During the first ten minutes of work, performance decrement is always most noticeable but then an asymptote is reached since there is little change afterwards. This has been the general finding in studies on vigilance and self-paced serial performance in the laboratory and also in various more realistic tests of long term performance (e.g. Riemersma et al. 1977). Traditional explanations of the decrement have included decreasing “arousal”, habituation, and also rising fatigue over time (Broadbent 197 1) but none is really satisfactorily because of ( 1) the very short-term on which the effect occurs, (2) its persistence over successive sessions, and (3) the finding that the decrement is often very small or even negligible (Teichner 1974). In fact, TOT appears to have less effect as subjects are more practiced (Bills 1931; Sanders and Hoogenboom 1970) which does not easily fit any of the explanations without extra provisions. As an interesting alternative, Buckner (1963) has proposed that instead of being “ underaroused” during the later period of work, subjects may be “extra aroused” in the beginning. In support of this suggestion, O’Hanlon (1970) found evidence for elevated sympathic activity during the first period of work in a vigilance test, which subsequently returned to a normal level, rather than an initially normal level that subsequently fell below average. In terms of the model of fig. 1 this suggests extra energetical supply through effort which rapidly wanes as time proceeds, until an asymptote is reached where the energetical supply reflects the states of arousal and activation without a major additional effort component. It is not possible to exclude incidental and variable contributions of effort at the later workperiods but as a first approximation this may be neglected. The main rationale is that, if effort still had a substantial contribution, then one would expect a further decrement of performance as a function of time. The case in point is that effort correction of suboptimal arousal and activation is
86
A. F. Sanders / Towards a model of stress and human performance
supposed to be a temporary matter unless there is continuing motivational stimulation. This view can also explain that the decrement, as observed in vigilance laboratory tests, is often not obtained in everyday work settings where subjects may find no reason to invest extra effort in the beginning of a workperiod (Riemersma et al. 1977). The notion that investment of effort is only a temporary means of adjusting energetical supply is also implicit in Broadbent’s (1971) upper-arousal concept, which is quite similar to the present effort notion. In fact Broadbent also suggested a hippocampal locus for upper-arousal and used the concept as a post-hoc explanation for the earlier discussed difficulties of the unidimensional arousal theory. The temporary character of the performance adjustment is presumably due to a lack of knowledge of the evaluation system about the quality of performance. In the absence of this knowledge, performance criteria might be rapidly eased. If, however, there is constant feedback or if there are other ways of keeping motivation high, the operation of effort should continue throughout the workspell. This element of the theory is confirmed by the well documented effect of KR, through which decrements of performance can be completely avoided. This is not only the case when considering the relatively small decrements observed as a function of TOT and time-of-day (e.g. Blake 1971), but also when there are sizeable decrements as in the case of sleep loss (Wilkinson 1969). KR is usually very effective in counteracting a decrement. A main argument to ascribe the effect of KR to effort and not to a direct increase of arousal or activation is that KR usually restores performance to the optimum. If performance happens to be at the optimum, KR has no or almost no effect (e.g. Wilkinson 1961). This would not be expected if KR operated on arousal and activation since that would lead to overarousal or overactivation and thus negatively affect performance. Although this has been occasionally found - e.g. in the interaction between the effects of noise and KR (Wilkinson 1963) the usual finding is that the optimum is restored by KR. In passing it is mentioned that restoration to the optimum argues against the common unwarranted notion that performance can be always simply improved by allocating extra effort. As advocated, allocation of effort is only useful if supply through arousal or activation deviates from optimum 141.
A. F. Sanders / Towards a model of stress and human performance
5.4. The effects of suboptimal
87
conditions revisited
The previous discussion was meant to clear the ground for the original question about the loci of the effects of suboptimal conditions. With regard to sleep state, it is worthwhile noting the usual interaction with the effects of TOT (e.g. Wilkinson 1961). Usually there is no effect of sleep loss during the first few minutes in laboratory tests whereafter a rapid rise is observed. This is illustrated in fig. 3 which is taken from the data of experiment 1 of Sanders et .al. (1982). There appeared a rapid increase of the effect of sleep loss followed by an approach to an asymptote wh.en subjects suffered from 24 hrs sleep loss and reacted to degraded signals. In the undegraded conditions the same trend was found but on a much smaller scale. There was not effect of TOT when subjects had slept normally. Similar relations between the effects of sleep state and TOT were obtained by Frowein (1981) with time uncertainty as an additional interacting variable. Together they suggest an effect of sleep loss on both arousal and activation which is compensated by effort during the first ;Jetiod of work. Wilkinson’s (196 1) finding that the effect of 24 hrs sleep loss completely disappeared when continuous KR is provided fits this interpretation. The pattern of results does neither indicate a direct adverse effect of sleep loss on computational processes nor on effort allocation. Most subjects appeared completely capable of allocating compensatory effort after 24 hrs sleep loss. The strong dependence on motivational factors as well as the predominant effect on long RT
[4] There is the recurring view in the literature that performance measures are basically unreliable indicators of mental functioning since performance always depends on the amount of effort invested (e.g. Jahns 1973). This notion is thought to be in line with Kahneman’s (1973) effort theory according to which a more “difficult” task consumes more effort than an “easy” task. Hence a difficult and an easy task may show about the same level of performance as long as the difference in difficulty is compensated by more effort investment. This view has been challenged by more recent capacity,theories (Navon and Gopher 1979; Lane 1977). The present model considers it also as unwarranted since effort has not the meaning of aspecific “processing capacity” but serves only the limited role of keeping the energetical supply to an optimum. Once the optimum is realized, performing the tasks depends fully on the computational processing requirements involved. It should be noted that Kahneman (1973) has also maintained that little is gained by trying harder, which similarly denounces the idea that performance is unrestrictedly determined by effort investment. Again, effort plays only a role in suboptimal conditions, so that questions concerning effort allocation can be only meaningfully studied by varying conditions which can be reasonably supposed to affect the energetical state.
A.F. Sanders / Towards a model of stress and human performance
88
prohibits an explanation of the effect of sleep state in terms of computational mechanisms. With regard to barbiturates the weight of the evidence is also towards an effect on a basal energetical mechanism - i.e. arousal although, due to a lack of research on relations with motivational variables, neither of the two alternatives can be firmly rejected. One of the regular findings is that effects of TOT and a small dosis of barbiturate are additive, which might, however, be due to the small size of the barbiturate effect and to the presence of proper placebo conditions. Together these factors might have precluded extra effort allocation in order to compensate the effect of the barbiturate in the same way as observed with sleep loss. They do not imply an adverse effect of barbiturates on effort (e.g. Trumbo and Gaillard 1975; Gaillard and Trumbo 1976). Again, the finding that the size of the barbiturate effect is strongest under time pressure and at the highest deciles of the
I
I
&
1
1
2
3
L
period Fig. 3. The effects of sleep state (sleep: dotted; sleep loss: drawn), signal degradation, signal-response compatibility and time-on-task (four five-minute periods) on choice reaction time.
A. F. Sanders / Towa&
a model of stress and human performance
89
distribution of reaction times argues against a direct effect on computational processes involved in the choice reaction test (Frowein and Sanders 1978). Yet more evidence - e.g. the relation with KR - is needed. This is also the case for the loci of the effect of amphetamines. Frowein ( 198 1, 198 1a) found small but reliable interactions between the effects of amphetamine and some motor variables under otherwise optimal conditions. From there he argued in favor of an activation effect of amphetamine. Also from the general literature he derived support for the view that amphetamine mainly affects motor processes. On the other hand, like KR, amphetamine has primarily the effect of counteracting a suboptimal state rather than affecting optimal conditions. This argues in favor of an effect on effort, which can also explain the negative relation between effects of amphetamine and response choice type of variables. According to fig. 1 too much supply of effort would disorient central processing. A key experiment in deciding whether effort is affected by amphetamine would be a study on the relation between the effects of sleep loss, signal quality and amphetamine. If only the basal activation level is raised, amphetamine will not compensate the effect of sleep loss on responding to degraded signals. If, on the other hand, effort is enhanced by amphetamine a compensation of this effect is predicted. It should be noted that amphetamine might have separate points of application to both effort and activation so that the two interpretations are not mutually exclusive. 5.5. Immediate arousal Another line of interest for the theory is contained in studies on immediate arousal evoked by loud auditory and tactual signals. The basic phenomenon is that when subjects are relatively unprepared to react - as with high time uncertainty about the arrival of a signal - a loud auditory signal or a tactual signal speeds up reaction time but only if the reaction task asks for simple or highly compatible reactions. In the case of incompatible S-R relations the effects of time uncertainty and auditory signal intensity turn out to be additive (Sanders and Andriessen 1978). This phenomenon has been interpreted as an arousing effect of a loud auditory or a tactual signal. The rise in arousal triggers a signal to the activation system, enhancing response readiness (see also Niemi 1979) and response execution along the upper loop in
90
A. F. Sanders / Towards a model of stress and human performance
fig. 1. Thus, the computational stages are bypassed and therefore there is the danger of rapid but badly considered reactions. This does not harm in the case of simple reactions but it is harmful in the case of stronger demands on response choice. In order to avoid unacceptable error rates with incompatible choice reactions, immediate arousal should be suppressed and this may be accomplished by uncoupling both systems through effort (Sanders 1980b). This has the consequence that the effects of signal intensity and time uncertainty become additive, as would be expected when considering a typical input and output variable (Sanders 1980b). Additional support for the immediate arousal hypothesis stems from the work of v.d. Molen and Keuss (1979, 1981) who found that increases in auditory intensity reduce RT to simple reactions but prolong RT to choice reactions, and in particular as they are more incompatible. 5.6. Robustness of the stage structure The immediate arousal studies show a second order interaction between the effects of time uncertainty, signal intensity and S-R compatibility, which does not make sense when explained by way of the additive factor logic. The reason is clearly that the upper loop of fig. 1 violates the assumption of single dimensional processing. Therefore the immediate arousal phenomenon is beyond application of the additive factor logic. Indeed, the second-order interaction might be considered as counterevidence for the logic as such. Elsewhere (Sanders 1980b) it has been argued .that such violations are only permissible if one has a satisfactory theoretical rationale as an explanation of the violation. With regard to immediate arousal the model of fig. 1 clearly provides this rationale by way of the upper loop which can be activated by strongly stimulating arousal through external stimuli. A similar rationale is not easily available for drug and sleep state effects and, therefore, if these variables would have affected the stage structure in the same way as immediate arousal, serious damage would have been inflicted to the theory. Fortunatelyi neither sleep nor drug state changed the mutual relations between the effects of the computational variables. This encourages further tests of the model and suggests that the experiments under review used conditions which either fitted the additive factor constraints or produced results which could be meaningfully related to the linear stage approach.
A.F. Sanders / Towards a model of stress and human performance
5.7.
Relations
91
to strategical effects of “stresses”
Is it possible to relate the present considerations to those in the context of resource strategy as discussed by Rabbitt (1979) and Hockey (1979)? If the present view is worth anything then such relations should be possible as a demonstration that principles derived from linear stages can be profitably combined with those obtained in situations where the additive factor methodology does not apply. A first strategical effect, described by Hockey (1979), concerns a trade-off between storage and throughput of information, mainly based on studies on the effects of noise and sleep state on tasks with presumably strong and weak memory components. In conditions of lack of sleep one tends to be less responsive and more fit towards storage, while in noise one tends more towards immediate throughput of information. Although the relation between sleep loss and storage has also been challenged (e.g. Spijkers 1980), it follows from the additive factor studies that after sleep loss subjects are indeed less responsive - perceptually as well as motoric - albeit this may well be the result of a deficient energetical supply and not of a change in strategy. There is little work on noise in relation to the model of fig. 1 but it sounds fair to assume a major effect of noise on arousal besides direct effects on perception proper (Poulton 1977). If noise has the effect of increasing supply it might first speed up the computational processes but, subsequently, at too high a level it might indeed lead to premature stage outputs and, hence, to errors. As a second principle Hockey (1979) discusses the interesting possibility of levelling versus sharpening of attentional resource allocation in conditions of respectively loss of sleep and noise. This principle is mainly based on results suggesting changes in cue utilization which seems more amorphous after sleep loss and which tends to be overselective during noise. Relevant cues are badly discriminated from irrelevant perceptual cues in the case of sleep loss. In contrast, attention is supposed to be “over-selective” in the case of an arousing condition, like noise. Apart from ignoring the irrelevant cues, noise would have the effect of ignoring relevant cues so that performance suffers. Hockey discusses examples from perceptual as well as from response and choice dominated tests. So far the linear stage research suggests one striking example of diminished cue utilization in the strong effect of sleep loss on identifying degraded signals. In fact, the identification of degraded
92
A. F. Sanders / Towards a model of stress and human performance
signals requires separating relevant from irrelevant cues in a way which is extremely close to Easterbrook’s (1959) original description. The linear stage model would suggest, however, that the results of Hockey and coworkers on perceptual selectivity have another energetical background than those on selectivity in choice. The last type of effect should be related to effort rather than to arousal. These two examples, preliminary and incomplete as they are, suggest the feasibility of relating the results derived from linear stage and resource strategy notions; at the same time it helps delineating their adequate areas of application. Resource models need linear stage models to propose interesting types of cognitive constraints. Linear stage models need resource models to avoid rigidity and provincialism. 5.8. Predictions and plans concerning stress With several recent suggestions (e.g. O’Hanlon 1981) the model of fig. 1 shares the deduction that subjects, showing high levels of performance under suboptimal conditions, are most under stress. They keep allocating effort to arousal and/or activation thus counteracting a decrement or a low performance asymptote. In particular if effort fails or is continually loaded, stress is high. Obviously, this poses a problem for interpreting nil results in research on suboptimal conditions. Did the suboptimal condition not affect arousal or activation or was there an effect that was compensated by effort? The problem is very similar to that of interpreting effects in terms of effort or arousal and activation (see p. 84). The solution is along the same line therefore: in research a condition should be created causing a low level of performance. Then, it should be examined whether the effect is erased by motivators like KR, pacing speed, social pressure, monetary rewards, or personality type. In case of full compensation a nil result is obtained that can be safely ascribed to effort. That is the very situation which is relevant to stress: it represents continuous loading of the effort mechanism which by definition leads to stress responses. The consequence is that performance measures are no proper indicants for stress, at least not in this type of approach. They are only needed as a control measure, ascertaining that sufficient effort is allocated to keep performance at the optimum. The proper stress response must be measured in physiological and hormonal patterns as reflection of the overdemands on effort.
A. F. Sanders / Towards a model of stress and human performance
93
Problems of underarousal and underactivation and their combined occurrence are most easily investigated by way of this technique. The psychological experiment is notoriously boring and, hence, an excellent setting for these types of stress. Even when limiting the discussion to these types, it is interesting to find out whether stable cognitive-energetical physiological patterns can be traced. Effortful compensation of activation may be accompanied by Mason’s hormonal pattern 1 - i.e. increased corticosteroid excretion but no elevated level of adrenaline. In contrast, compensation of a low level of arousal should at least be accompanied by effects of catecholamines. Arousal is supposed to be related to adrenaline but probably more in the case of overstimulation (Frankenbaeuser et al. 1971). The immediate arousal paradigm might be a satisfactory situation for studying this possibility. In general, however, conditions of overarousal and overactivation and the hypothetical attempts towards damping their effect are quite hard to create. The traditional approaches do not suffice and ethical constraints are rapidly reached. Yet, it may not be altogether impossible witness the study of Ursin et al. (1978) on student paratroopers. From a performance point of view it would be highly interesting to test the present model under conditions of tension and anxiety. In those tests, stress is unrelated to the experimental task but its effects on those tasks are examined. Through proper combination with cognitive variables, relating to arousal and activation the stress situations could be interpreted in terms of the basal mechanisms. In addition the extent of effort compensation can be examined. These situations are really needed to study the proposed stress patterns. From that study it is still a considerable step to an inquiry about the relation between stress patterns and the various illnesses which are traditionally ascribed to stress, but, again, there could well be systematic relationships.
References Appley, M.H. and R. Trumbull, 1967. ‘On the concept of psychological stress’. In: M.H. Appley and R. Trumbull (eds.), Psychological stress. New York: Appleton. Bainbridge, L., 1978. Forgotten alternatives in skill and workload. Ergonomics 21, 169-185. Bartlett, F.C., 1953. ‘Psychological criteria of fatigue’. In: W.F. Floyd and A.T. Welford (eds.), Symposium on fatigue. London: Lewis. Berlyne, D., 1969. ‘The development of the concept of attention in psychology’. In: C.R. Evans and T.B. Mulholland (eds.), Attention in neurophysiology. New York: Appleton.
94
A. F. Sanders / Towards a model of stress and human performance
Bills, A.G., 1931. Blocking: a new principle in mental fatigue. American Journal of Psychology 43, 230-245. Blake, M.J., 1971. ‘Temperature and time-of-day’. In: W.P. Colquhoun (ed.), Biological rhythms and human performance. New York: Academic Press. Buckner, D.N., 1963. ‘An individual-difference approach to explaining vigilance performance’. In: D.N. Buckner and J.J. McGrath (eds.), Vigilance: a symposium. New York: Academic Press. Broadbent, D.E., 1971. Decision and stress. London: Academic Press. Broadbent, D.E. and M.H.P. Broadbent, 1980. ‘Priming and the passive/active model of word recognition’. In: R.S. Nickerson (ed.), Attention and performance, 8. Hillsdale, NJ: Erlbaum. Cannon, W.B., 1927. The James-Lange theory of emotion. American Journal of Psychology 39. 106-139. Comstock, E.M., 1973. Processing capacity in a letter-matching task. Journal of Experimental Psychology 100, 63-72. Cox, T., 1978. Stress. London: Macmillan. Donders, F.C., 1969. On the speed of mental processes. (Translation by W.G. Koster.) [Originally published: 1868.1 Acta Psychologica 30, 412-431. Duffy, E., 1962. ‘Activation’. In: N.S. Greenfield and R.A. Stembach (eds.). Handbook of psychophysiology. New York: Holt. Easterbrook, J.A., 1959. The effect of emotion on cue utilisation and the organisation of behavior. Psychological Review 66, 183-20 I. Egeth, H., J. Jonides and S. Wall, 1972. Parallel processing of multielement displays. Cognitive Psychology 12, 278-286. Frankenhaeuser, M., B. Nordheden, A.L. Myresten and B. Post, 1971. Psychophysiological reactions to understimulation and overstimulation. Acta Psychologica 35, 298-308. Frowein, H.W., 198la. Selective drug effects on human information processing. Soesterberg: Institute for Perception TNO. Frowein, H.W., 198lb. Selective effects of barbiturate and amphetamine on information processing and response execution. Acta Psychologica 47, 105- 115. Frowein, H.W., 1981~. ‘Effects of two counteracting stresses on the reaction process’. In: A.D. Baddeley and J.L. Long (eds.), Attention and performance, 9. Hillsdale. NJ: Erlbaum. Frowein, H.W. and A.F. Sanders, 1978. Effects of amphetamine and barbiturate in a serial reaction task under paced and self-paced conditions. Acta Psychologica 42, 263-276. Frowein, H.W., A.W.K. Gaillard and CA. Varey, 1981. Evoked potential compounds. visual processing stages and the effects of a barbiturate. Biological Psychology 13, 239-249. Gaillard, A.W.K., i978. Slow brain potentials preceding task performance. Amsterdam: Academische Pers. Gaillard, A.W.K., 1980. ‘The use of task variables and brain potentials in the assessment of cognitive impairment’. In: B.M. Kulig et al. (eds.). Epilepsy and behavior. Lisse, The Netherlands: Swets and Zeitlinger. Gaillard, A.W.K. and D.A. Trumbo, 1976. Drug effects on heart rate and heartrate variability during a prolonged reaction task. Ergonomics 19, 6 1 l-622. Gopher, D. and A.F. Sanders, 1983. ‘S-Oh-R: Oh stages! Oh resources!’ In: W. Prinz and A.F. Sanders (eds.), Cognition and motor behavior. Heidelberg: Springer. Gottsdanker, R., 1975. ‘The attaining and maintaining of preparation’. In: P.M.A. Rabbitt and S. Domic (eds.), Attention and performance, V. London: Academic Press. Gray, J., 1971. The psychology of fear and stress. London: Weidenfeld and Nicolson. Hamilton, P., G.R.J. Hockey and J.G. Quinn, 1972. Information selection, arousal and memory. British Journal of Psychology 63, I8 1 - 189. Hamilton, P., G.R.J. Hockey and R. Reyman, 1977. ‘The place of the concept of activation in human information processing’. In: S. Domic (ed.), Attention and performance, 6. Hillsdale, NJ: Erlbaum.
A. F. Sanders / Towar& a model of stress and human performance
95
Hebb, D.O., 1955. Drives and the conceptual nervous system. Psychological Review 62, 243-254. Hockey,’ G.R.J., 1979. ‘Stress and the cognitive components of skilled performance’. In: V. Hamilton and D.M. Warburton (eds.), Human stress and cognition. New York: Wiley. Hull, C.L., 1943. Principles of behavior. New York: Appleton-Century-Crofts. Jahns, D.W., 1973. A concept of operator workload in manual vehicle operations. Report 14. Forschungs Institut fur Anthropotechnik, Meckenheim, Germany. Kahneman, D., 1973. Attention and effort. Englewood Cliffs, NJ: Prentice-Hall. Kimble, D.P., 1968. Hippocampus and internal inhibition. Psychological Bulletin 70, 285-295. Ktilpe, 0.. 1895. Outlines of psychology. New York: Macmillan. La Berge, D., 1975. ‘Acquisition of automatic processing in perceptual and associative learning’. In: P. Rabbitt and S. Domic (eds.), Attention and performance, V. London: Academic Press. Lacey, J.I., 1967. ‘Somatic response patterning and stress: Some revisions of activation theory’. In: M.H. Appley and R. Trumbull (eds.), Psychological stress: some issues in research. New York: Appleton-Century-Crofts. Lane, D.M., 1977. Attention allocation and the relationship between primary and secondary task difficulty. A reply to Kantowitz and Knight. Acta Psychologica 41, 493-495. Lazarus, R.S., 1966. Psychological stress and the coping process. New York: McGraw-Hill. Logan, G.D., 1978. Attention in character classification: evidence for the automaticity of components stages. Journal of Experimental Psychology: General 107, 32-63. Logan, G.D., 1979. On the use of a concurrent memory load to measure attention and automaticity. Journal of Experimental Psychology: Human Perception and Performance 5, 189-207. Logan, G.D., 1980. Attention and automaticity in Stroop and priming tasks: theory and data. Cognitive Psychology 12, 523-533. Lynn, R., 1966. Attention, arousal and the orientation reaction. Oxford: Pergamon. Mackay, C. and T. Cox, 1979. Response to stress: occupational aspects. Guilford: IPC Science and Technology. Mackworth, J.F., 1964. Performance decrement in vigilance, threshold, and high-speed perceptual motor tasks. Canadian Journal of Psychology 18, 209-223. Mackworth, J.F.. 1969. Vigilance and habituation. Middlesex: Penguin. Malmo, R.G., 1959. Activation: a neuropsychological dimension. Psychological Review 66, 367-386. Mason, J.W., 1975. ‘Emotion as reflected in patterns of endoctrine integration’. In: L. Levi (ed.), Emotions, their parameters and measurement. New York: Raven. McClelland, J.L., 1979. On the time relations of mental processes: an examination of systems of processes in cascade. Psychological Review 86, 287-330. Mirski, A.F. and C. Kometski, 1964. On the dissimilar effects of drugs on the digit symbol substitution and continuous performance tests. Psychopharmacologia 5, 161-177. v.d. Molen, M.W. and P.J.G. Keuss, 1979. The relationship between reaction time and intensity in discrete auditory tasks. Quarterly Journal of Experimental Psychology 31, 95- 102. v.d. Molen. M.W. and P.J.G. Keuss, 1981. Response selection and the processing of auditory intensity. Quarterly Journal of Experimental Psychology 33, 177-184. Monk, T.H. and V.C. Leng, 1982. Time of day effects in simple repetitive tasks. Acta Psychologica 51. 207-221. Moray, N., 1967. Where is capacity limited? A survey and a model. Acta Psychologica 27, 84-92. Ntitanen, R., 1973. ‘The inverted-U relationship between activation and performance: a critical review’. In: S. Komblum (ed.), Attention and performance, 4. New York: Academic Press. NZ&&nen, R. and A. Merisalo, 1977. ‘Expectancy and preparation in simple reaction time’. In: S. Dornic (ed.), Attention and performance, 6. Hillsdale, NJ: Erlbaum. Navon, D., 1977. Forest before trees: the precedence of global features in visual perception. Cognitive Psychology 9, 353-383.
96
A. F. Sanders / Towarrls a model of stress and human performance
Navon, D. and D. Gopher, 1979. On the economy of human information processing systems. Psychological Review 86, 214-255. Niemi, P., 1979. Stimulus intensity effects on auditory and visual reaction processes. Acta Psychologica 43, 299-3 12. Norman, D.A. and D.G. Bobrow, 1975. On data-limited and resource-limited processes. Cognitive Psychology 7, 4464. G’Hanlon, J.F., 1970. Vigilance, the plasma catecholamines. and related biochemical and physiological variables. Techn. Rep. 787-2. Goleta, CA: Human Factors Research. Inc. O’Hanlon, J.F., 1981. Boredom: practical consequences and a theory, Acta Psychologica 49. 53-82. Pachella, R.G., 1974. ‘The interpretation of reaction time in information processing research’. In: B.H. Kantowitz (ed.), Human information processing. Hillsdale, NJ: Erlbaum. Posner, MI., 1978. Chronometric explorations of mind. Hillsdale, NJ: Erlbaum. Posner, M.I., 1980. Orienting of attention. The Quarterly Journal of Experimental Psychology 32. 3-26. Posner, M.I. and S.J. Boies, 1971. Components of attention. Psychological Review 78, 391-408. Posner, MI. and C.R.R. Snyder, 1975. ‘Facilitation and inhibition in the processing of signals’. In: P.M.A. Rabbitt and S. Domic (eds.), Attention and performance, 5. New York: Academic Press. Poulton, E.C., 1977. Continuous intense noise masks auditory feedback and inner speech. Psychological Bulletin 84, 977- 1001. Pribram, K.H. and D. McGuinness, 1975. Arousal, activation and effort in the control of attention. Psychological Review 82, 116-149. Proctor, R.W., 1978. Attention and modality-specific interference in visual short-term memoq. Journal of Experimental Psychology: Human Learning and Memory 4. 239-245. Rabbitt, P.M.A., 1979. ‘Current paradigms and models in human information processing’. In: V. Hamilton and D.M. Warburton (eds.), Human stress and cog&ton. New York: Wiley. Riemersma, J.B.J., A.F. Sanders, C. Wildervanck and A.W.K. Gaillard, 1977. ‘Performance decrement during prolonged night driving’. In: R.R. Mackie (ed.), Vigilance. New York: Plenum. Routtenberg, A., 1968. The two-arousal hypothesis: reticular formation and the limbic system. Psychological Review 75, 5 I-80. Salthouse, T.A., 1981. Converging evidence for information-processing stages: a comparative influence stage-analysis method. Acta Psychologica 47, 39-61. Sanders, A.F., 1971.. Probabilistic advance information and the psychological refractory period. Acta Psychologica 35, 128-137. Sanders, A.F., 1972. Foreperiod duration and the time course of preparation. Acta Psychologica 36, 60-71. Sanders, A.F., 1977. ‘Structural and functional aspects of the reaction process’. In: S. Domic (ed.). Attention and performance, 6. Hillsdale, NJ: Erlbaum. pp. 3-25. Sanders, A.F., 1979. ‘Some effects of instructed muscle tension on choice reaction and movement time’. In: R. Nickerson (ed.), Attention and performance. 8. Hillsdale, NJ: Erlbaum. Sanders, A.F., 1980a. Stress, activatie en verrichtingen. Nederlands Tijdschrift voor Psychologie 35, 185- 199. Sanders, A.F., 1980b. ‘Stage analysis of reaction processes’. In: G. Stelmach and J. Requin (eds.). Tutorials on motor behavior. Amsterdam: North-Holland. Sanders, A.F., 1981. ‘Stress and human performance: a working model and some applications’. In: G. Salvendi and M.J. Smith (eds.), Machine pacing and occupational stress. London: Taylor and Francis. Sanders, A.F. and J.E.B. Andriessen, 1978. A suppression effect of response selection on immediate arousal in a choice reaction task. Acta Psychologica 42, 18 I- 186.
A. F. Sanders / Towardr a model of stress and human performance
91
Sanders, A.F. and A.A. Bunt, 1971. Some remarks on the effect of drugs, lack of sleep, and loud noise on human performance. Nederlands Tijdschrift voor Psychologie 26, 670-684. Sanders, A.F. and W. Hoogenboom, 1970. On the effects of continuous active work on performance. Acta Psychologica 33, 414-431. Sanders, A.F. and W.D. Reitsma, 1982a. The effects of sleep-loss on processing information in the functional visual field. Acta Psychologica 51, 149-162. Sanders, A.F. and W.D. Reitsma, 1982b. Sleep-loss and covert orienting of attention. Acta Psychologica 52, 137-145. Sanders, A.F., J.L.C. Wijnen and A.E. v. Arkel, 1982. An additive factor analysis of the effects of sleep-loss on reaction processes. Acta Psychologica 51, 41-59. Selye, H., 1936. A syndrome produced by diverse nocuous agents. Nature 138, 32-34. Selye, H., 1956. The stress of life. New York: McGraw-Hill. Selye, H., 1979. ‘The stress concept and some of its implications’. In: V. Hamilton and D.M. Warburton (eds.), Human stress and cognition. New York: Wiley. Shiffrin, R.M., 1975. ‘The locus and role of attention in memory systems’. In: P.M.A. Rabbitt and S. Dornic (eds.) Attention and performance, 5. New York: Academic Press. Smith, E.E., 1968. Choice reaction time: an analysis of the major theoretical positions. Psychological Bulletin 69, 77- 110. Sperandio, J.C., 1972. Charge de travail et regulation des processes operatoires. Travail Humaine 35, 85-98. Stanger, R., 1977. Homeostasis, discrepancy, dissonance: a theory of motives and motivation. Motivation and Emotion 1, 103- 138. Stemberg, S., 1969. On the discovery of processing stages: some extension of Donders’ method. Acta Psychologica 30, 276-315. Spijkers, W., 1980. Slaapdeprivatie en taakprestatie: een literatuuroverzicht. Nederlands Tijdschrift voor Psychologie 35, 151-172. Taylor, D.A., 1976. Stage analysis of reaction time. Psychological Bulletin 83, 161-191. Teichner, W.H., 1974. The detection of a simple visual signal as a function of time of watch. Human Factors 16, 339-353. Trumbo, D.A. and A.W.K. Gaillard, 1975. ‘Drugs, time uncertainty, signal modality and reaction time’. In: P.M.A. Rabbitt and S. Dormic (eds.), Attention and performance, 5. New York: Academic Press. pp. 44-454. Ursin, H., E. Baade and S. Levine, 1978. Psychobiology of stress: a study of coping men. New York: Academic Press. Warburton, D.M., 1979. ‘Physiological aspects of information processing and stress’. In: V. Hamilton and D.M. Warburton (eds.), Human stress and cognition. New York: Wiley. Welford, A.T., 1967. Single channel operation in the brain. Acta Psychologica 27, 5-22. Welford, A.T., 1973. Stress and performance. Ergonomics 16, 567-580. Wilkinson, R.T., 1961. Interaction of lack of sleep with knowledge of results, repeated testing and individual differences. Journal of Experimental Psychology 62, 263-27 1. Wilkinson, R.T., 1963. Interaction of noise with knowledge of results and sleep deprivation. Journal of Experimental Psychology 66, 332-337. Wilkinson, R.T., 1969. Some factors influencing the effects of environmental stressors upon performance. Psychological Bulletin 72, 260-272. Yerkes, R.M. and J.D. Dodson, 1908. The relation of strength of stimulus to rapidity of habit formation. Journal of Comparative Neurology and Psychology 18, 459-482.