Emerging Needs and Opportunities for Human Factors Research (1995)

J. Frank Yates, Roberta L. Klatzky, and Carolynn A. Young

The media have given considerable attention to the stress in contemporary society. For instance, the topic was the cover theme of a Time magazine issue in 1983 (June 6). Within the past few years, there have been numerous highly publicized incidents in which U.S. Postal Service employees have attacked and killed their supervisors and coworkers. The "headline" segment on a recent ABC News 20/20 program was devoted to these events. The program speculated that job stress, including the stress associated with the Postal Service's efforts to improve its highly automated sorting systems, was a significant contributor to the violence.

The current focus on stress extends beyond the popular press. Scholarly analyses of several prominent tragedies have cited stress as a factor:

• Wickens (1992) conjectures that stress may have impaired the performance of a control room operator on duty during the nuclear reactor accident at Three Mile Island, making the situation worse than it might have been otherwise. Wickens suggests that stress may have played a similar role in the downing of an Iranian airliner by the crew of the U.S. missile frigate Vincennes, who thought they were under attack by a military aircraft.

• In March 1989, Air Ontario Flight 363 crashed and burned on takeoff from Dryden, Ontario. The pilot, copilot, a flight attendant, and 21 passengers were killed. The immediate cause of the accident was snow and

ice on the wings. Helmreich (1990) performed a human factors study of the incident. He concluded that time pressure and other stressors quite likely induced the pilots to make faulty—and ultimately fatal—decisions about deicing and departure.

• In every armed conflict, military personnel sometimes mistakenly fire on their own confederates. There were numerous instances of such "friendly fire" among American troops in the Persian Gulf War of 1991. Military analysts acknowledge a host of causes of these incidents, including cognitive ones like poor situational awareness (U.S. Army, 1992a, 1992b); these causes are also thought to include factors commonly acknowledged as forms or concomitants of stress, such as anxiety.

The stress literature is not restricted to high-profile occurrences like those just described. There is considerable published work on the stresses of everyday life, including—or perhaps especially—in the ordinary workplace (Manuso, 1983). In 1988, the National Research Council published a highly influential report by its Committee on Techniques for the Enhancement of Human Performance (Druckman and Swets, 1988). The committee did a critical review of research on a host of performance-related topics, ranging from motor skills to paranormal phenomena. Subsequent work by the committee (Druckman and Bjork, 1991) extended the range of topics even further. The 1988 report (p. 115) acknowledged that "none of the topics … has received more attention than the management of stress." This impression seems accurate; the scholarly literature on stress is enormous. The American Psychological Association's PsycINFO database lists over 35,000 entries under the keyword stress. Using the joint keyword query stress and performance, a recent search directed toward topics more closely associated with human factors yielded 2,565 entries.

Everyone seems to acknowledge that stress is a serious problem, and the research community has appeared to respond by devoting a great deal of energy to solving that problem. Why, then, is the topic being proposed as an area of additional emphasis in human factors research? There are several responses to this question, which we will develop more fully below. Briefly, however, the thrust of our argument is not that we need to increase the sheer volume of stress research (although that may well be true), but that future research should direct more attention to particular cognitive aspects of stress-performance interactions, aspects that have managed to escape close scrutiny. (By convention, the term cognition refers to acts of perception and knowing.) A hint of this oversight is suggested by a search of the PsycINFO database using the query stress and cognition. This search produced only 223 items, a yield of fewer than 9 items per year over the 1967-94 coverage of the database. In their comprehensive survey, Mross

and Hammond (1989) have also noted the sparseness of the stress-cognition literature.

A good case also can be made that stress and the significance of its consequences will increase over the coming years:

• Global economic competition has increased dramatically over the past couple of decades, and it shows little sign of slackening in the near future. Thus, both public organizations, such as the U.S. Postal Service, and private organizations will undoubtedly find themselves severely challenged simply to survive. These challenges are virtually guaranteed to cause every person in those organizations to experience greater stress. For instance, polls suggest that there is an emerging sense among the public that job security is rarer now than in the past and that this condition will last a long time (Church, 1993).

• Economic and technological developments have led to greater interdependence among individuals and organizations, often accompanied by centralization. This implies that an action by any element in the contemporary workplace has wider—and hence more serious—consequences than in the past (see also Driskell and Salas, 1991b). Take the case of financial markets. On a dollar volume basis, the bulk of the activity in these markets is undertaken by the managers of large institutions, such as pension funds, not by individual investors (Siconols, 1992). Hence, as the market crash of 1987 demonstrated, buying and selling by a remarkably small number of people can have dramatic and far-reaching effects. So, to the extent that stress influences those individuals' choices, its ramifications extend far beyond their personal welfare and affect most people.

• The potential effects of stress on cognition are further amplified by the changing nature of work itself. In almost every arena, today's jobs place greater emphasis on cognitive than on motor performance. For example, as noted by Hartzell (1992), at one time a good helicopter pilot had to exercise extraordinary manual control. These days, however, a good pilot depends more on the ability to select appropriate procedural sequences, which in turn are executed by computer-controlled devices.

• There is also the role of new technology itself. For instance, it is now possible for employers to continuously monitor the performance of workers such as telephone operators and telemarketers. There is some evidence that this practice itself exacerbates the stress experienced by workers (e.g., Rogers et al., 1990).

Hans Selye (1956) is the scientist who first called special attention to the concept of stress in humans. Selye (1983) noted that the body's physiological

reactions to an injury or intense emotion can be divided into those that are peculiar to the particular stimulus in question and those that appear in response to any noxious stimulus. He called the latter constellation of responses the general adaptation syndrome (GAS). It includes three stages (Selye, 1983:4-5): (a) an ''alarm reaction," representing a "general call to arms of the body's defensive forces," including changes in body temperature, blood pressure, and hormonal secretions; (b) a "stage of resistance," in which there is adaptation to the stressor and the reduction or even disappearance of the elements of the alarm reaction; and (c) a "stage of exhaustion," when the compensatory measures in the resistance stage deplete the available resources. Selye essentially equated stress with this nonspecific physiological response pattern. For some time, most work on stress and psychological functioning continued to emphasize this biological, response-oriented characterization of stress. Thus, for instance, Broadbent (1963) and others sought to account for stress effects in terms of fluctuations in arousal.

The concept of stress has evolved over the years. Stress researchers are far from unanimous in their use of the term (see, e.g., Everly and Sobelman, 1987:Chapter 1). And this situation has prompted at least some authors to suggest that the very concept of stress has outlived its usefulness (e.g., Hammond and Doyle, 1991). Nevertheless, there is reasonable consensus about the meaning of the expression in work on cognition, and we will interpret it here according to that consensus (see Hancock and Warm, 1989). Therefore, by stress we mean an individual's reactions to apparent significant threats to his or her welfare, reactions that often entail heightened emotion (see Keinan et al., 1987; Yates, 1990). As articulated by Novaco (1988:4), a threat is a self-perceived imbalance between the demands made on a person and the resources he or she can apply to satisfying those demands. That is, the person doubts his or her ability to meet the challenge imposed by the circumstances. Several implications of this stress conception can be highlighted with the aid of Figure 10.1.

To start with, stress is a response to situational conditions. If these conditions induce the perception of threat, they achieve the status of stressors—i.e., "that which causes stress" (Selye, 1983:9).

The personal circumstances element in Figure 10.1 calls attention to the fact that identical conditions can have different threat implications at different times. That is, a given individual's varying resources, physical and otherwise, alter that person's vulnerability. A problem-solving task that is daunting when one is exhausted at day's end is "no sweat" after a good night's sleep.

The personal characteristics feature highlights two kinds of stable individual differences. First are the differences in people's abilities to respond to challenges. Second are differences in the intensity with which people experience (or anticipate) the losses that would ensue should they fail to

FIGURE 10.1 A schematic representation of stress and its connections with other entities.

meet a challenge. These differences are well illustrated in the discussion of medical devices used in the home in Chapter 4. Even a trained outpatient is likely to see himself or herself as less competent at using the equipment than a professional and is thus more likely to experience stress in an emergency. And since the patient's own health or life—not someone else's—is at stake, stress should be even more intense.

The personal appraisals element in Figure 10.1 is an acknowledgment that ignorance or special sensitivity to situational conditions can preclude or exacerbate the experience of threat (Lazurus and Folkman, 1984; Paterson and Neufeld, 1987). As an extreme case, if a surgical attendant is completely oblivious of a cardiac monitor's malfunctioning, he or she will perceive no threat to the patient's survival and hence experience no stress.

Finally, the other factors part of Figure 10.1 is a recognition that, in virtually any situation, the cognitive activities in question will be affected by other things besides stress. And the effects of these factors may or may not be independent of stress effects. Lighting quality has an obvious influence on machine assembly performance. But within bounds, that influence may or may not differ for people who are subjected to large rather than small amounts of stress.

The conceptual framework depicted in Figure 10.1 is minimal. Nevertheless, it does provide a useful perspective for the ensuing discussion, and it will be elaborated in the context of that discussion.

Human factors as a discipline focuses on the interactions between the people and the artifacts that together compose human-technology systems

for accomplishing various tasks. Thus, a nuclear power plant has a human factors problem if displays of reactor status induce high probabilities of operator error, and an anesthesiology protocol has poor human factors features if it encourages physicians and technicians to administer inappropriate dosages of gas. Typical activities that people undertake in human-technology systems have both cognitive and noncognitive aspects. Implicit in the earlier discussion is the assumption that stress often has adverse effects on the cognitive elements of those activities. This is a common and plausible assumption, but there may be conditions under which stress has beneficial influences on cognition. Indeed, several theories imply what some of those conditions might be (e.g., Easterbrook, 1959; Yates, 1990:Chapter 13). We thus submit that a major theme in human factors research should be deepening our fundamental understanding of stress-cognition relationships. A complementary aim should be the development of practical techniques for counteracting negative effects and exploiting positive ones.

Consider an experiment by Rothstein (1986). On any given trial in that experiment, the subject's task was to predict a criterion variable, C. The subject was shown vertical bars labeled A and B. The heights of these bars served as numerical cues for C, which was statistically related to A and B in a particular way. For instance, in one condition, the optimal rule for predicting C might have the form

where f and g are specific linear or inverted-U functions. Via a block of subject-controlled trials with feedback, each subject first learned to predict the criterion to a satisfactory level of accuracy. Half of the subjects then performed test trials under time pressure. On each trial, the subject had to predict C from A and B within a deadline of six seconds. One of Rothstein's conclusions was that time pressure induced the subjects to apply more widely differentiated weights to the cues. So, if a subject initially tended to predict C according to a rule similar to

then time pressure altered that rule to one more like

The tentative generalization was that time pressure tends to make judgment policies less evenhanded.

Rothstein's (1986) results are interesting in themselves. But we highlight his experiment here because in one important respect it is typical of laboratory studies of stress and cognition. Recall that the consensus definition of stress requires a person to doubt his or her ability to meet a task demand. Missing-response data suggest that Rothstein's six-second time limit did indeed engender such doubt. However, another stress requirement is that the affected person should recognize task failure as a significant threat to his or her welfare. For average college students (as were Rothstein's subjects), failing to complete any assigned intellectual task is probably an ego threat. But in Rothstein's study, that seemed to be the only plausible threat; subjects were simply "required" (p. 85) to respond within the time limit. Therefore, in terms of threat, the stressfulness of Rothstein's time pressure was probably minimal. Rothstein did not explicitly apply the term stress to his time pressure conditions. Nevertheless, his manipulations were actually quite representative of how researchers have often sought to induce stress in controlled settings.

The problem illustrated by this example is that the stressors used in laboratory studies are essentially benign (almost a contradiction in terms); they constitute almost no threat to the subject. From an ethical, as well as a strictly legal, perspective (in the United States, at least), this is how it should and must be. After all, no responsible researcher would want to risk causing the serious harm implicit in the "significant threat" element of the stress construct.

Standard informed consent protocols have another consequence, too. Not only must the potential consequences of a subject's actions be mild; the conditions confronting the subject cannot be surprising. As numerous stress specialists (e.g., Levine, 1988) have noted, uncertainty, including complete unexpectedness, is a critical element in many truly stressful situations. As a fanciful example, consider the popular science fiction film Alien. Perhaps the film's most terrifying moment occurs shortly after the protagonists conclude that their embryonic extraterrestrial tormentor had inexplicably but surely disappeared, leaving them safe and secure. Then, with shocking suddenness, they learn that the monster has been with them all along. It had been maturing parasitically within a crew member's body. Everyone would agree that the crew's stress level—to say nothing of the audience's—is then much higher than had the alien been the usual, garden-variety, Godzilla-type monster.

Most experimentalists are surely aware that the stress levels they induce in the laboratory are far milder than those in the real-world situations to which they wish to generalize. So, implicit in their work is the hopeful assumption that, although the effects of laboratory stressors are markedly

weaker than those of severe natural stressors, they are qualitatively the same. But not all stress investigators are so optimistic. Take the situation in which the stakes in a real-world decision problem can be a significant stressor in themselves (Janis and Mann, 1977). The prospect of losing a 50-cent bonus because of a bad decision in a laboratory experiment might seem so inconsequential that the subject can treat the decision task as simply an exercise in pure reasoning, on a par with solving a math problem. But suppose the situation entails the possibility of losing a daughter's life because of choosing the wrong medical treatment (see Ritov and Baron, 1990). Mann (1992) speculates that the very nature of the decision process is fundamentally different in this kind of distinctly stressful circumstance.

Right now it is impossible to say whether what we have learned from laboratory studies of stress and cognition does, in fact, generalize to more extreme levels of the stressors involved. But, as was implied above, we do know that those experiments can tell us virtually nothing about the role of uncertainty in stress effects. There is yet another limitation: the fact that experiments are short-term affairs. Acute stressors are present only briefly, whereas chronic stressors are sustained for extended periods (e.g., on long space flights or submarine tours of duty). For the most part (there are exceptions, of course), human factors and basic cognition researchers have experimentally examined the contemporaneous effects of acute stressors on, for instance, task performance in the presence of noise or extreme temperatures (see Hancock, 1986b). In contrast, clinical psychologists and other health care providers (e.g., Newberry et al., 1987) have more often addressed the long-term consequences of extended stress exposure, consequences that might occur in contexts completely removed from the locus of the stressor. Post-traumatic stress syndrome, such as combat fatigue, is perhaps the best-known example (Levine, 1988). As suggested by the work of Cohen (1980; Cohen and Spacapan, 1978), the effect of chronic stress can present surprises. These surprises seem inaccessible via standard laboratory techniques.

Given the above observations, it is essential that human factors investigators of stress and cognition seek to create ethical yet effective means of studying the influences of stressors that more closely approximate those in real-world human-technology systems. Indeed, it is plausible that a major reason for the very sparseness of the stress and cognition literature is that researchers often find it prohibitively difficult to create appropriate conditions.

How might this methodological research challenge be approached? Numerous avenues could be explored. One point of departure is to note that the issues are similar to those that confront investigators in education and medicine. In these fields, the stakes can be the difference between a child's academic success and failure or even between life and death. (Should this child participate in a new, experimental learning method? Should this patient

receive a new drug or a placebo?) Methodologists and ethicists in education and medicine have responded to the challenge with numerous innovations (e.g., Cook and Campbell, 1979). Human factors researchers would do well to see whether some of those innovations can be adapted to the demands of stress research. At a minimum, human factors methodologists might find it instructive to examine the process by which their counterparts in education and medicine have addressed analogous dilemmas.

It is impossible to anticipate which, if any, research strategies developed in education and medicine warrant close examination for emulation in stress and cognition research. However, a good case can be made for investigating the feasibility of methods that entail central roles for two specific techniques: (a) the analysis of naturally occurring incidents and (b) the study of simulations, including competitive games.

Helmreich's (1990) inquiry into the Air Ontario crash at Dryden is a good illustration of incident analysis. The investigator sifted through the available records of a significant, catastrophic event, trying to reconstruct conditions that may have led or contributed to it. Among those conditions were ones that fit prevailing stress theories. It so happened that, because of his previous work, Helmreich was aware of stress effects and hence would have been on the lookout for their possible involvement in the Dryden incident. This might not always be the case whenever incident analyses are routinely performed (e.g., in high-risk domains such as aviation, nuclear power, and surgery). We therefore recommend that protocols for regularly commissioned incident analyses (e.g., by the Nuclear Regulatory Commission, the Federal Aviation Administration, and various surgery review boards) be designed so that the possible role of stress can be evaluated critically.

Among the inherent weaknesses of full-blown mishap analyses are their expense and the (fortunate) rareness of the incidents. One variation on the incident analysis approach relies on statistical explorations of richer (but necessarily less detailed) routine archival data and could thus circumvent these drawbacks. An example of the requisite kind of information source is the database maintained by the National Aeronautics and Space Administration's Aviation Safety Reporting System (ASRS). This is an archive of reports submitted by individuals throughout the U.S. civilian aviation system. Such reports, which are recorded anonymously, describe any conditions or activities the reporter thinks might be hazardous. Williams et al. (1992) successfully analyzed ASRS reports about incidents involving resource management and geographic disorientation. There is nothing to preclude similar analyses of stressed-related incidents.

Another variation on incident analysis entails embedding experiments or quasi-experiments (Cook and Campbell, 1979) within naturally occurring activities that are likely to be stressful, but to which certain individuals have already voluntarily committed themselves. One example (which was not considered particularly successful by the investigators themselves) was a study by Idzikowski and Baddeley (1983) of the reactions of inexperienced public speakers prior to delivering colloquium papers. Another widely cited pair of studies were those in which Fenz (1974; Fenz and Jones, 1972) traced arousal patterns in parachutists preparing to make jumps. The embedding studies that have perhaps been least susceptible to subject self-selection biases are those that tested ways of teaching surgery patients to manage the stress associated with their impending operations (e.g., Langer et al., 1975).

Incident analyses have been performed for years (see, e.g., Woods, 1993). The reports resulting from such analyses, as well as the manuals that sometimes guide them, suggest a remarkable degree of methodological sophistication and thoroughness. For example, the National Transportation Safety Board's Investigator's Manual prescribes in great detail who is to participate in the investigation of airplane crashes and what their roles should be. An especially attractive feature of the manual is a checklist of human performance factors that should be examined; for example, life habit patterns, training, and control design. There has been some discussion of critical issues underlying incident analyses (e.g., Reason, 1990). Nevertheless, it appears that the scholarly literature on such issues is surprisingly scant. The rarity of relevant reports highlights both an opportunity and a need for methodological developments to which human factors stress researchers could be major contributors.

In collaboration with practicing incident analysts from diverse back-grounds, human factors specialists should aggressively seek to extend the public literature on incident analysis techniques. There are numerous specific issues such a literature should address, as our illustrations have highlighted. Take the case of Helmreich's (1990) examination of the Air Ontario crash at Dryden. We conjectured that Helmreich's awareness of stress theories probably sensitized him to the possible role of stress in the crash. But critics might say that such awareness actually threatens the integrity of incident analyses. For example, research has shown (e.g., Chapman and Chapman, 1967) that clinical psychologists who hold theories that certain signs ought to be indicative of patients' true conditions tend to see "illusory correlations" between those signs and conditions even when in reality no connections exist. A strong literature on incident analysis methods would address techniques for precluding such validity threats.

In any simulation, the aim is to duplicate some, but not all, of the essential features of a particular experience (see Jones et al., 1985). For instance, flight simulators are used in pilot training mainly because they are much cheaper to build and operate than real aircraft; they closely approximate the behavior of an aircraft but not its costs. The present research recommendation is that the same approach be attempted in the domain of stress and cognition. That is, an effort should be made to develop stress simulators as research tools that would induce realistic stress reactions but would not actually expose subjects to the threat of serious harm that is an essential element of stress.

Passable stress simulators should in principle be more difficult to construct than physical simulators. History has shown that a functional flight simulator needs to emulate only the physical response to an operator's manipulation of a control device (including what is seen through the cockpit windshield). The details of actually building such a simulator might be quite involved, but the underlying physical (and perceptual) principles are well known. The builder of a stress simulator, however, must bring about psychological illusions; the subject should feel stress without actually being subject to the hazards that are normally a prerequisite for it. Unfortunately, what we know about stress principles is not enough to guide us in creating those illusions in a systematic way. Thus, to the extent that stress simulators are feasible, they must emanate from researchers' intuitions and everyday experiences.

One starting point for developing stress simulators might be commonplace activities that appear to already approximate what we would like the simulators to do. Several examples come to mind: (a) films, (b) amusement park rides, (c) role-playing exercises, and (d) competitive games. We can all recall encounters with one or more of these activities in which we experienced what felt like serious stress. (Think of your most memorable horror movie, fun house, or roller coaster ride.) At one level of consciousness, we were fully cognizant that we were in no real danger; otherwise, we would never have agreed to participate. However, that assessment was temporarily suspended and put completely out of mind. Effectively, we allowed ourselves to be encapsulated within a small, insular world where gut-wrenching threats abounded.

It is unclear exactly what makes an effective horror movie, for instance. And perhaps that is the genius of artistry. But it should be possible to systematically review good and poor films, rides, role-playing exercises, and the like in order to glean hints of techniques that could be mimicked in building useful stress simulators. For example, one plausible working hypothesis

is that a good simulator must insulate the subject from external stimulation that would encourage thoughts like, "This is only a game," "I realize that nothing bad can actually happen to me," or "In the larger scheme of things, this just doesn't matter."

In the abstract, the kinds of stress simulators suggested here seem almost far-fetched or frivolous. They are not, however. Instantiations of similar ideas already exist and have proven their value. Take the case of role-playing. Armstrong (1987) has reviewed research (including some of his own) about forecasting the outcomes of conflict situations, such as union negotiations. He specifically compared the relative accuracy of forecasts generated by expert opinion and those that relied on role plays of the conflict situations. Forecasts based on role-playing tend to be far better.

Or take competitive sports. There is an active subarea of sports psychology that focuses on stress and performance among athletes (see Jones and Hardy, 1990). Work in this area should be examined not only for methodological ideas but also for possible generalizations of its substantive findings beyond the domain of sports.

There is currently considerable activity in the development of virtual reality technology. Much of that work is proceeding at a breakneck pace in the private entertainment industry, to the point that several virtual reality games and "rides" are currently available to consumers in special arcades and amusement parks (Corliss, 1993). One perspective on these developments is that they are expected to be so lucrative precisely because they promise to be so much better than competing technologies (e.g., conventional video games and films) at simulating extreme stressors. We suggest that their potential as stress and cognition research platforms be thoroughly explored. Should the anticipated commercial success of virtual reality entertainment devices be realized, researchers might also find themselves with a ready-made volunteer subject pool analogous to the skydivers studied by Fenz (1974) and his colleagues. In fact, researchers might even want to explore developing research simulators that are in fact "games" included among other games in commercial arcades and parks. Decision researchers have sometimes tested the generalizability of behavioral decision-making principles in the high-stakes, real-world environment of Las Vegas casinos (e.g., Goodman et al., 1979; Lichtenstein and Slovic, 1973).

One paradoxical complication of stress simulation should be acknowledged and addressed by stress research methodologists and ethicists. Our discussion has emphasized the simulation of stress without the real danger of direct physical harm. For instance, in a simulator, there is no chance of a subject's being killed in a high-speed crash. But to the extent that real stress is induced, there is still the potential for psychological and indirect physical harm from the stress itself (see, e.g., Cox, 1988). In the vernacular,

the simulation might "scare the subject out of his wits." Screening subjects on appropriate risk factors is one seemingly reasonable approach to such possibilities (e.g., Shanteau and Dino, 1993).

We have noted that existing stress and cognition theories suggest that there are conditions in which stress can be expected to enhance performance. However, the overarching theme of our recommendations for substantive research is that human factors investigators should develop practical means for counteracting or handling adverse stress-cognition interactions. Five approaches to solving human factors stress problems can be distinguished:

Approach 1: Eliminate or weaken the stressor.

Approach 2: Reduce stress reactions by the current system participants.

Approach 3: Select system participants who are stress resistant if not stress immune.

Approach 4: Train system participants to function effectively even when stressed.

Approach 5: Design system features and procedures so that system goals are achieved despite the presence of high stress levels among system participants-—stress-proofing, as it were.

To varying degrees, all five of these approaches are discernible in the stress-related literature (most of which does not explicitly address human factors). Nevertheless, some of them—approaches 1 and 2, in particular—are emphasized more than others. In fact, approaches 1 and 2 can be seen as the standard strategies for dealing with stress. The underlying idea is that if stress is a bad thing, get rid of it. In approach 1, the work situation is designed to minimize the presence and intensity of stressors, for instance, by eliminating noise or altering workload schedules (see Chiles, 1982). In approach 2, an effort is made to change how system personnel react to stressors. A classic example of this approach is "stress inoculation training (SIT)" (Meichenbaum, 1985). When a person is inoculated against an infectious disease (e.g., chicken pox), he or she is exposed to mild forms of the disease with the expectation that the body will develop the ability to resist the more severe natural forms of that disease. Similarly, a key element of SIT is that an individual is exposed to progressively more intense forms of a stressor and is assisted in developing tools for managing that stressor, to the point that stress reactions no longer occur.

The logic of stress-elimination approaches is not unreasonable. However, as noted by Druckman and Swets (1988), it is impossible to eliminate

or even reduce stress all the time. No matter what we do, some crises will inevitably occur. We should not simply concede defeat and allow nature (in this instance, dysfunctional stress-cognition interactions) to take its course. Approaches 3 to 5 are specific tacks for engineering effective system performance no matter what circumstances might arise. We will now make research recommendations for each of these strategies.

Implicit in the conceptual scheme of Figure 10.1—and consistent with common experience—is the assumption that individuals tend to differ in how stressful they find particular circumstances and how well they perform under those conditions. Also as suggested by the conceptual scheme, part of the reason for this variation is that people have different competencies. Thus, Operator A is less stressed by task demands than Operator B because he or she knows the job better; it constitutes no threat. The more interesting situation is that in which equally competent individuals differ in their stress reactions. Our conceptual scheme implies that this could occur for either of two reasons. First, these individuals might appraise situations and their competencies differently; the stressed person could see tasks as more demanding and his or her skills as weaker, thereby anticipating less adequate performance. Alternatively, individuals might vary in the intensity with which they feel the pain of task failure. Such appraisal and reactivity differences are a major focus of work on decision and motivation processes (see Arkes, 1982; Yates, 1990). Consider, for example, why not all consumers choose the same car or why not all students pursue their studies vigorously.

Fortunately, stress and cognition researchers have begun seeking to understand the nature of individual differences and the bases for them (see Hockey et al., 1986, especially Section IV: ''Individual Differences, Adaptation, and Coping"). However, for our purposes, the essential issues are somewhat different. In particular, the primary question is whether and how we can accurately and economically predict which individuals are likely to be unaffected by severe stressors in given situations.

Consider air traffic control. To the lay observer, the responsibilities and the pace of air traffic control seem crushingly stressful. Airplanes appear to converge on airports unceasingly, and a single mistake could be fatal for hundreds of people. There is some evidence that air traffic control specialists (ATCSs) do exhibit unusually high stress symptoms (e.g., Cobb and Rose, 1973), but the literature is by no means unanimous on this conclusion (Crump, 1979; Smith, 1985). Numerous studies indicate that ATCSs do experience stress levels comparable to those of other workers but they regard stress as among the least of their concerns. It is conceivable—in

fact, plausible—that the stress resistance of the individuals who compose the corps of working ATCSs is higher than that of the initial pool of trainees. Somehow, because of self-selection and dismissals for correlated factors, stress-prone trainees leave the program. This natural filtering process might be an effective way to choose stress-resistant participants for a system in which stressors are inevitable. However, that process is undoubtedly expensive and perhaps risky and inefficient. It thus seems appropriate that a priority for stress and cognition researchers is to explore other ways to identify people who are stress resistant.

Do people, in fact, exhibit behaviors that could be used to predict stress resistance? Perhaps so. For instance, there are indications that ATCSs are reliably distinct from the general population with respect to several easily assessed personal characteristics, such as low trait anxiety and tendencies toward nonconformity (Smith, 1985). Research in the more general personality literature reveals other encouraging signs (but see Driskell and Salas, 1991b). For example, Allred and Smith (1989) have shown that individuals who score high on so-called hardiness scales tend to find pleasant and inviting challenges in what others regard as unpleasant stressors (see also Kobasa, 1979). Parkes (1986) similarly found that extraversion scores among nurses were correlated with their reactions to work stressors. Finkelman and Kirschner (1980) have reported data consistent with their proposal that laboratory measures of information processing, such as delayed digit recall, could be used as specific predictors of stress resistance in human-technology contexts such as air traffic control. And work reviewed by Reason (1988) indicates that stress vulnerability is correlated with simple questionnaire reports of everyday "cognitive failures." We previously recommended the development of various forms of stress simulators, such as competitive games and virtual reality experiences. If that effort succeeds, then it seems only natural that performance in the simulations be examined for their ability to predict individuals' performance in the work situations of ultimate interest: if trainee Smith thinks clearly in tight game situations, is it reasonable to expect him or her to do the same when a catastrophe strikes the workroom?

Keinan and Friedland (1984) performed an interesting experiment in which Israeli military personnel served as subjects. The focal task required the subject to search a page for randomly dispersed instances of a designated digit, say, "3." At the end of the session, the subject's skill at the task was evaluated by performance on three criterion trials. (Surprisingly, there are actually better and worse ways of performing such tasks.) After each of the first two test trials, the subject received a 1.5 mA shock to his

fingers. The five conditions in the experiment were distinguished according to the context in which the subject was trained to perform the task. During high-fidelity training, the subject received a 1.5 mA shock after 4 of the 10 training trials. In low-fidelity training, these shocks were of 0.2 mA intensity. Shocks of 0.2, 0.75, 1.0, and 1.5 mA intensity were administered in succession during graduated-intensity training. Shocks at these same levels of intensity were administered in random sequence during random-intensity training. Finally, subjects in a no-stress condition received no shocks during training. Figure 10.2 shows the mean numbers of correct target identifications, by condition, during the three criterion trials and the last three training trials. We recommend that a priority for future research be the development of procedures for training personnel to perform well even under stressful conditions (see also Means et al., 1993). Keinan and Friedland's results speak directly to how such development might proceed.

First, note that the performance of the group trained under no stress was superior to that of any other group. Moreover, the performance of the graduated-fidelity and random-intensity subjects was especially poor. This suggests that stress can be expected to interfere significantly with skill acquisition (but see Christianson, 1992, for a contrasting view). Moreover,

FIGURE 10.2 Mean number of correct target identifications as a function of training session stress. SOURCE: Keinan and Friedland (1984:Table 1).

it seems that the worse the stress, the worse the interference. (Consistent with previous discussions about stress effects, the apprehension and uncertainty in the graduated-fidelity and random-intensity conditions should amplify the associated stress.) The implied conclusion is that personnel should be trained to perform required tasks under relaxed conditions. This contrasts with the strategy behind the various stress conditions in the Keinan and Friedland experiment. In that view, training should be most effective when it occurs under conditions similar to those in which trainees must eventually perform.

Keinan and Friedland's results indicate that we should expect a relaxed-training approach to yield some performance decrement when a stressor is eventually encountered. (Observe that criterion performance was worse than terminal training performance for the no-stress group.) But there might be a way to preclude this decrement. A plausibly effective hybrid training strategy would train in basic skills under relaxed conditions and then allow trainees to "overlearn" those skills under stressful conditions. Yet another potentially effective strategy would be to concede that skill acquisition is slower under stress, but insist that it simply needs to be extended for a longer period of time. Neither of these strategies was tried directly by Keinan and Friedland, but these and others should be high on the research agenda.

Leaving aside the use of electrical shock, research like that of Keinan and Friedland is clearly the logical starting point for the kind of work that is required to inform the development of techniques for training people to perform well under stress. But as we have suggested, this work needs to be extended in various ways. There are also numerous matters of detail. The Keinan and Friedland experiment addressed only one training issue: When and in what form should stress be introduced into a training regimen? Consider some of the other issues, which are closely tied to fundamental questions about stress and cognition.

Visual search similar to that required of Keinan and Friedland's subjects is clearly an important part of many duties in human-technology systems. However, there are many other activities in which such well-defined tasks play a minimal role. In fact, we can expect that in crisis situations (e.g., during catastrophic weather conditions or equipment breakdowns) the required tasks will be anything but well defined; instead, they will be free form in the extreme. As was argued above, the inherent uncertainty in such unique situations can be expected to intensify the stress that is experienced (Harris, 1987). A close examination of the stress literature suggests that almost all experimental stress research has studied performance on well-defined

tasks. Therefore, our beliefs about stress influences, including those indicated by Keinan and Friedland's results, might not be as generalizable as we would like. Future studies should determine whether or not this is the case.

The notion of controlled versus automatic information processing is widely recognized (Schneider and Shiffrin, 1977; Shiffrin and Schneider, 1977). This notion is different from the distinction between strategic and reactive stress responses but is related to it. In a strategic response, the individual executes a deliberate plan for coping with stressful conditions. For instance, faced with a stringent time limit for making a choice, a decision maker might rationally elect to alter the decision process, giving temporal priority to those aspects of the options that are deemed most important (e.g., Ben Zur and Breznitz, 1981). The decision maker realizes that there might be insufficient time to deliberate all the factors he or she would ideally weigh and thus concludes that he or she must ignore some of them (see also Payne et al., 1988). In contrast, a reactive response is automatically evoked by a stressor; the individual cannot help himself or herself. An example is suggested by Janis and Mann's (1977) theory of decision behavior. This theory proposes that the nature of the decision process is conditioned by the intensity of the stress involved. When stress is at its zenith, the theory suggests that the decision maker enters a stage of hypervigilance characterized in part by the kind of loss of control associated with pure reactivity (see also Janis et al., 1983).

It seems that the nature of stress performance training should be dictated by whether a stressful situation tends to evoke strategic or reactive responses. If responses tend to be strategic, then we might want to focus on things like incentives for selecting one type of action rather than others. Such tactics would be ineffective when responses are normally reactive. In the latter case, we would have to "reprogram" the subject to spontaneously perform different actions given appropriate conditions, a process that must rely on seemingly countless repetitions. A good analogy is what occurs in sports training. Tennis coaches sometimes have their students repeat a maneuver over and over again. That way, when the corresponding conditions occur during a tense match, the player does the right thing without having to think; it just happens. Thus, part of the stress training research agenda should be directed toward distinguishing which stress responses are strategic and which are reactive. A related issue that must be addressed is deciding which responses an organization might want to be strategic rather than reactive, or vice versa.

Exactly what should personnel be trained to do when performing their duties under stress? Suppose that an operator is instructed to do A, then B, and then C under normal conditions. On one hand, we might expect stress to induce the operator to deviate from the A-B-C routine when he or she ought to adhere to it. If so, training must be directed to maintaining that adherence. But how? On the other hand, an emergency might require that ideally the operator act differently from what would be appropriate under normal conditions. How should the operator be taught to respond most adequately to the altered circumstances? To the best of our knowledge, these kinds of essential issues have not been directly addressed, and clearly they must be. One way researchers might proceed is to begin with what previous research has taught us about adverse stress-cognition interactions. Stress-performance training protocols can then be pointedly designed to counteract those interactions, as we will illustrate below.

We noted previously—and hence this call for more research—that the stress and cognition literature has always been sparse. Nevertheless, that literature indicates consensus about several fundamental principles that we hope will be verified by future work (see, e.g., Hamilton and Warburton, 1979; Hockey, 1983, 1986; Hockey and Hamilton, 1983). For instance, there seems to be general agreement that stress tends to reduce short-term memory (STM) performance. It is not clear how and why this occurs (see also Christianson, 1992). There might be a reduction in sheer capacity. Alternatively, capacity might remain constant, but must now be divided between a focal task and other tasks, such as monitoring the stressor (Cohen, 1978; Reason 1988). It is important for stress researchers to determine the basis for this decrement in STM performance. Suppose that worsened STM performance is due to attention being shared with monitoring and managing the stressor. It should be easier to recover the attention if that stressor is an ambient stressor than if it is a task stressor. An ambient stressor (e.g., heat, noise, close confinement) is one whose presence is independent of the person's task performance, whereas the opposite is true of a task stressor (e.g., time pressure or high performance stakes) (Yates, 1990).

Regardless of the basis for the reduction of STM performance with stress, there are direct training implications. For example, suppose a task analysis indicates that an operator's duties require him or her to perform mental operations that rely heavily on STM (e.g., operations like those required in the decision algorithms Zakay and Wooler, 1984, trained their subjects to perform). Then the operator should be taught to distrust his or her ability to carry out those operations unaided when under duress; instead, the operator might use paper and pencil and double-check.

Another consensus conclusion is that stress reduces the scope of perceptual

attention (Baddeley, 1972), a phenomenon sometimes called tunneling. Thus, stress reduces the range of elements in the environment a person sees or hears. An adjustable-beam spotlight provides a useful metaphor for perceptual narrowing. The claim is that stress reduces the width of the beam. A plausible alternative hypothesis is that the width of the beam remains constant but the beam simply develops a more erratic sweep (see Keinan, 1987; Keinan et al., 1987).

As in the case of STM, there are several fundamental issues that need to be resolved. Assuming that perceptual narrowing does indeed occur, along what gradient does it proceed? As noted by Yates (1990), in the context of decision making, at least three different gradients have been proposed in the literature. In the Easterbrook (1959) hypothesis, attention is restricted according to the objective, functional significance of various items of information for decision quality. The thesis implicit in the work of Ben Zur and Breznitz (1981), however, says that attention is reduced according to the personal evaluative significance of various features of the choice alternatives. And then there is the idea that (in one form, originated in drive theory) stress reduces attention according to the inverse of the probabilities of information being taken into account under nonstressful conditions (see Zajonc, 1968). Another attention-narrowing question concerns memory. Much of the information used in cognitive tasks is retrieved from memory. Does stress induce retrieval-narrowing phenomena analogous to perceptual restrictions?

The ultimate resolution of fine-grained theoretical issues like perceptual-narrowing gradients will indeed have practical implications. Nevertheless, stress training designers need not wait for a complete understanding of the theoretical issues. As suggested in the case of STM decrements, trainees might be warned that perceptual narrowing is likely to occur. But if the experience of debiasing in judgment and decision making is a guide (see Fischhoff, 1982), this is unlikely to decrease inappropriate narrowing to any great extent. Instead, the trainee must be given specific prescriptions and must be taught when and how to apply them, or perhaps even "conditioned" to implement them essentially automatically. Exactly what those prescriptions might be will depend on the designers' ingenuity.

There have been some indications that stress alters people's reliance on information from different sources. For instance, Wickens et al. (1989) cite evidence that stress encourages people to depend more heavily on information retrieved directly from long-term memory (LTM), presumably at the expense of information that is perceived directly from the environment or synthesized anew in STM (see also Klein, 1989). It is as if stress increases the significance of a person's prior experiences. Perhaps this is why, independently of competency differences, stress is sometimes found to have greater effects on performance for novices than for experts (Baddeley, 1972;

Hancock, 1986a). Greater dependence on LTM might be advantageous in some stressful situations. However, the kind of rigidity it implies might well be dysfunctional in crisis situations involving novel circumstances (e.g., a breakdown in a mechanism that has never broken down). There is at least some evidence that stress does in fact reduce certain forms of creativity that are required in such circumstances (Shanteau and Dino, 1993; Voss, 1977). Once again, training designers must be inventive in devising ways to counteract these effects.

Our final recommendation is similar to the previous one, but it applies to human-technology systems rather than to individual participants. Specifically, we suggest that researchers seek to develop system features and routines that allow the systems to achieve their goals even when participants are stressed. This should begin with an effort to understand how stressors affect both individual and group behavior. We have already discussed some of what is known about stress and cognition interactions at the individual level, including the fact that the literature is so sparse. For various reasons the literature on stress at the group level is even more limited (see Davis et al., 1992; Park, 1990). Thus, researchers who take up the challenge posed by this final proposal will need to make contributions to fundamental understanding of the problem as well as to practical innovations.

The reactions of the individual are key to the design of systems that function well in stressful circumstances. In our previous discussion, we suggested that training efforts be directed toward equipping system participants with personal skills to counteract the adverse effects of stress on cognition. The system could also be redesigned in a way that gives participants tools and procedures to counteract those effects. For example, recall that one effect of stress is to restrict the range of attention and consider crisis situations, such as the medical emergencies involving outpatients' operating sophisticated equipment discussed in Chapter 4. A specially constructed expert system might force operators to explicitly acknowledge that they have reviewed checklists presented to them by the system.

Or consider the creativity limitations we should expect in crises, when creativity is at a premium. Although the kinds of options that would occur to any single individual during stress are likely to be restricted, different individuals should be expected to generate somewhat different ideas. Thus, systems might be designed so that under stressful conditions, procedures require that inputs from multiple participants be elicited and synthesized. ("Two heads are better than one," and so on.) Not surprisingly, such activities can be done in ways that are more of a hindrance than a help (see Hill,

1982). Fortunately, however, there is a literature that discusses how such group "process losses" can be avoided (e.g., Dennis and Valacich, 1993).

As we have indicated, stress influences on group processes have been studied less extensively than we might like. Nevertheless, the literature that does exist suggests that stress is likely to alter communication and perhaps authority patterns among human-technology system participants and that not all of those changes are for the better (e.g., Driskell and Salas, 1991a; Gladstein and Reilly, 1985). For instance, Janis (1972) has proposed that stress is a major antecedent of the dysfunctional form of organizational decision making called "groupthink." In groupthink, a policy-making group becomes so preoccupied with achieving concurrence that other aspects of the decision task suffer from sometimes-fatal neglect. As researchers attempt to develop the kinds of stress-resistant systems envisioned, they should include features that anticipate and circumvent such dysfunctional social patterns.

As we noted at the outset, there are numerous indications that the stresses experienced by the people involved in human-technology systems have intensified in recent years and are likely to continue doing so for some time in the future. Human roles in these systems increasingly place greater demands on cognitive than on, say, motor activities. This implies that good human factors design will increasingly rely on our understanding of the connections between stress and cognition. An examination of the literature reveals that the requisite knowledge base is surprisingly skimpy, given broad-based interest in stress. We concluded that a major contributor to this state of affairs is probably the inherent difficulty of inducing significant stress in subjects while protecting those subjects from the potential harm inherent in the stress construct.

Our analysis implied that a prerequisite for meeting the challenge is the development of better methods of doing stress and cognition research. Hence, our first class of research recommendations is methodological. Perhaps modeling their efforts on those of methodologists and ethicists in education and medicine, human factors investigators should seek to develop techniques that are practically and ethically feasible yet capable of revealing how significant stress does indeed affect cognition. Any and all promising approaches should be explored. But we argued that two especially promising methodological avenues should have priority. One of them relies on analyses of naturally occurring incidents involving stressful conditions, such as accidents. The other uses various forms of simulation, including competitive games and virtual reality technologies.

We concluded that the pressing needs for substantive research on stress

and cognition are most productively conceptualized in terms of alternative approaches for handling adverse effects of stress on cognitive performance. Three particular approaches seem to demand special attention:

• Selecting stress-resistant personnel. The critical assumption underlying this approach is that people differ reliably in their tendencies to perform cognitive tasks well or poorly under stressful conditions. The aim of the research recommended in this area is identifying easy-to-assess predictors of such individual differences.

• Training people for stressful conditions. A fundamental yet unresolved training issue that needs to be addressed is when and in what form stress should be introduced into training programs. Other recommendations address questions that need to be resolved in order to provide specific guidance to stress trainers: How, if at all, do stress effects on cognition depend on whether the cognitive task is well defined as opposed to free form? Which cognitive responses to stress are strategic and which are reactive, and what are the implications of the distinction? What specific countermeasures for such adverse stress responses as diminished short-term memory performance are effective—and why?

• Stress-proofing human-technology systems. Most discussions of stress have emphasized the behavior of individuals. However, a good case can be made that much is to be gained by designing human-technology systems, including protocols for social interactions, so that the systems function well even during times of heightened stress. Our broad recommendation is that human factors researchers seek out principles that would inform such design objectives. Specific recommendations call for the development of expert systems and organization schemes that compensate for anticipated human difficulties, such as restricted attention and reduced creativity.

The practical challenges posed by research on stress and cognition are formidable and have undoubtedly contributed to the paucity of such work in the past. However, there is reason to be optimistic that developments along the lines suggested here will meet with greater success than in the past. That work promises more than immediate practical benefits. It should also lead to significant advances in our understanding of fundamental principles of cognition and stress.

Allred, K.D., and T.W. Smith 1989 The hardy personality: cognitive and physiological responses to evaluative threat. Journal of Personality and Social Psychology 56:257-266.

Arkes, H.R. 1982 Psychological Theories of Motivation, 2nd ed. Monterey, Calif.: Brooks/Cole.

Armstrong, J.S. 1987 Forecasting methods for conflict situations. Pp. 157-176 in G. Wright and P. Ayton, eds., Judgmental Forecasting. Chichester, England: Wiley.

Baddeley, A.D. 1972 Selective attention and performance in dangerous environments. British Journal of Psychology 63:537-546.

Ben Zur, H., and S.J. Breznitz 1981 The effect of time pressure on risky choice behavior. Acta Psychologica 47:89-104.

Broadben, D.E. 1963 Differences and interactions between stresses. Quarterly Journal of Experimental Psychology 15:205-211.

Chapman, L.J., and J.P. Chapman 1967 Genesis of popular but erroneous psychodiagnostic observations. Journal of Abnormal Psychology 72:193-204.

Chiles, W.D. 1982 Workload, task, and situational factors as modifiers of complex human performance. Pp. 11-56 in E.A. Alluisi and E.A. Fleishman, eds., Human Performance and Productivity. Hillsdale, N.J.: Erlbaum.

Christianson, S.-A. 1992 Emotional stress and eyewitness memory: a critical review. Psychological Bulletin 112:284-309.

Church, G.J. 1993 Jobs in an age of insecurity. Time November 22:32-39.

Cobb, S., and R.M. Rose 1973 Hypertension, peptic ulcer, and diabetes in air traffic controllers. Journal of the American Medical Association 224:489-492.

Cohen, S. 1978 Environmental load and the allocation of attention. Pp. 1-29 in A. Baum, J.E. Singer, and S. Vallins, eds., Advances in Environmental Psychology. Hillsdale, N.J.: Erlbaum. 1980 Aftereffects of stress on human performance and social behavior: a review of research and theory. Psychological Bulletin 88:82-108.

Cohen, S., and S. Spacapan 1978 The aftereffects of stress: an attentional interpretation. Environmental Psychology and Nonverbal Behavior 3:43-57.

Cook, T.D., and D.T. Campbell 1979 Quasi-Experimentation: Design and Analysis Issues for Field Settings. Chicago, Ill.: Rand McNally.

Corliss, R. 1993 Virtual, man! Time 1 (November):80-83.

Cox, T. 1988 Psychobiological factors in stress and health. Pp. 603-628 in S. Fisher and J. Reason, eds., Handbook of Life Stress, Cognition and Health. New York: Wiley.

Crump, J.H. 1979 Review of stress in air traffic control: its measurement and effects. Aviation, Space, and Environmental Medicine 50:243-248.

Davis, J.H., T. Kameda, and M.F. Stasson 1992 Group risk taking: selected topics. Pp. 163-199 in J.F. Yates, ed., Risk-Taking Behavior. Chichester, England: Wiley.

Dennis, A.R., and J.S. Valacich 1993 Computer brainstorms: more heads are better than one. Journal of Applied Psychology 78:531-537.

Driskell, J.E., and E. Salas 1991a Group decision making under stress. Journal of Applied Psychology 76:473-478. 1991b Overcoming the effects of stress on military performance: human factors, training, and selection strategies. Pp. 183-193 in R. Gal and A.D. Mangelsdorff, eds., Handbook of Military Psychology. Chichester, England : Wiley.

Druckman, D., and R.A. Bjork, eds. 1991 In the Mind's Eye: Enhancing Human Performance. Committee on Techniques for the Enhancement of Human Performance, National Research Council. Washington, D.C.: National Academy Press.

Druckman, D., and J.A. Swets, eds. 1988 Enhancing Human Performance: Issues, Theories, and Techniques . Committee on Techniques for the Enhancement of Human Performance, National Research Council. Washington, D.C.: National Academy Press.

Easterbrook, J.A. 1959 The effect of emotion on cue utilization and the organization of behavior. Psychological Review 66:183-201.

Everly, G.S., Jr., and S.A. Sobelman 1987 Assessment of the Human Stress Response. New York: AMS Press.

Fenz, W.D. 1974 Arousal and performance of novice parachutists to multiple sources of conflict and stress. Studia Psychologica 16:133-144.

Fenz, W.D., and G.B. Jones 1972 The effect of uncertainty on mastery of stress: a case study. Psychophysiology 9:615-619.

Finkelman, J.M., and C. Kirschner 1980 An information-processing interpretation of air traffic control stress. Human Factors 22:561-567.

Fischhoff, B. 1982 Debiasing. Pp. 422-444 in D. Kahneman, P. Slovic, and A. Tversky, eds., Judgment Under Uncertainty: Heuristics and Biases. New York: Cambridge University Press.

Gladstein, D.L., and N.P. Reilly 1985 Group decision making under threat: the tycoon game. Academy of Management Journal 28:613-627.

Goodman, B., M. Saltzman, W. Edwards, and D.H. Krantz 1979 Prediction of bids for two-outcome gambles in a casino setting. Organizational Behavior and Human Performance 24:382-399.

Hamilton, P., and D. Warburton, eds. 1979 Human Stress and Cognition. Chichester, England: Wiley.

Hammond, K.R., and J.K. Doyle 1991 Effects of Stress on Judgment and Decision Making, Part II. Final Report for Contract No. DAAL03-86-D-001. Research Triangle Park, N.C.: U.S. Army Research Office.

Hancock, P.A. 1986a The effect of skill on performance under an environmental stressor. Aviation, Space, and Environmental Medicine 57:59-64. 1986b Sustained attention under thermal stress. Psychological Bulletin 99:263-281.

Hancock, P.A., and J.A. Warm 1989 A dynamic model of stress and sustained attention. Human Factors 31:519-537.

Harris, J.H. 1987 Prior task experience and psychological stress: an investigation. Pp. 45-53 in J.H. Humphrey, ed., Human Stress: Current Selected Research , Vol. 2. New York: Ames Press.

Hartzell, J. 1992 Army-NASA Aircrew-Aircraft Integration (A³I) Program. Presentation (in October) to the Committee on Human Factors, National Research Council, Atlanta, Ga.

Helmreich, R.L. 1990 Human factors aspects of the Air Ontario crash at Dryden, Ontario: analysis and recommendations to the Commission of Inquiry. Pp. 319-348 in V.P. Moshansky, ed., Technical Appendices to the Final Report of the Commission of Inquiry Into the Air Ontario Crash at Dryden, Ontario.

Hill, G.W. 1982 Group versus individual performance: are N + 1 heads better than one? Psychological Bulletin 91:517-539.

Hockey, G.R.J., ed. 1983 Stress and Fatigue in Human Performance. Chichester, England: Wiley.

Hockey, G.R.J. 1986 Changes in operator efficiency. In K. Boff, L. Kaufman, and J. Thomas, eds., Handbook of Perception and Performance, Vol. II. New York: Wiley.

Hockey, R., and P. Hamilton 1983 The cognitive patterning of stress states. Pp. 331-362 in G.R.J. Hockey, ed., Stress and Fatigue in Human Performance. Chichester, England: Wiley.

Hockey, G.R.J., A.W.K. Gaillard, and M.F.H. Coles, eds. 1986 Energetics and Human Information Processing. Dordrecht, Netherlands: Martinus Nijhoff.

Idzikowski, C., and A. Baddeley 1983 Waiting in the wings: apprehension, public speaking and performance. Ergonomics 26:575-583.

Janis, I.L. 1972 Victims of Groupthink. Boston, Mass.: Houghton Mifflin.

Janis, I.L., and L. Mann 1977 Decision Making. New York: Free Press.

Janis, I.L., P. Defares, and P. Grossman 1983 Hypervigilant reactions to threat. Pp. 1-42 in H. Selye, ed., Selye's Guide to Stress Research, Vol. 3. New York: Scientific and Academic Editions.

Jones, J.G., and L. Hardy, eds. 1990 Stress and Performance in Sport. Chichester, England: Wiley.

Jones, E.R., R.T. Hennessy, and S. Deutsch, eds. 1985 Human Factors Aspects of Simulation. Working Group on Simulation, Committee on Human Factors, National Research Council. Washington, D.C.: National Academy Press.

Keinan, G. 1987 Decision making under stress: scanning of alternatives under controllable and uncontrollable threats. Journal of Personality and Social Psychology 52:639-644.

Keinan, G., and N. Friedland 1984 Dilemmas concerning the training of individuals for task performance under stress. Journal of Human Stress 10:185-190.

Keinan, G., N. Friedland, and Y. Ben-Porath 1987 Decision making under stress: scanning of alternatives under physical threat. Acta Psychologica 64:219-228.

Klein, G.A. 1989 Recognition-primed decisions. Advances in Machine Systems Research 5:47-92.

Kobasa, S.C. 1979 Stressful life events, personality, and health: an inquiry into hardiness. Journal of Personality and Social Psychology 37:1-11.

Langer, E.J., I.L. Janis, and J.A. Wolfer 1975 Reduction of psychological stress in surgical patients. Journal of Experimental Social Psychology 11:155-165.

Lazarus, R.S., and S. Folkman 1984 Stress, Appraisal, and Coping. New York: Springer.

Levine, S. 1988 Stress and Performance. Background paper prepared for the Committee on Techniques for the Enhancement of Human Performance, National Research Council, Washington, D.C.: National Academy Press. Available through the National Academy Press's Publication-on-Demand Program.

Lichtenstein, S., and P. Slovic 1973 Response-induced reversals of preference in gambling: an extended replication in Las Vegas. Journal of Experimental Psychology 101:16-20.

Mann, L. 1992 Stress, affect, and risk taking. Pp. 201-230 in J.F. Yates, ed., Risk-Taking Behavior. Chichester, England: Wiley.

Manuso, J. 1983 The Equitable Life Assurance Society program. Preventive Medicine 12:658-662.

Means, B., E. Salas, B. Crandall, and T.O. Jacobs 1993 Training decision makers for the real world. Pp. 306-326 in G.A. Klein, J. Orasunu, R. Calderwood, and C.E. Zsambok, eds., Decision Making in Action: Models and Methods. Norwood, N.J.: Ablex.

Meichenbaum, D. 1985 Stress Inoculation Training. New York: Pergamon.

Mross, E.F., and K.R. Hammond 1989 Annotated Bibliography for Cognition and Stress. Report No. 295. Center for Research on Judgment and Policy. Boulder: University of Colorado.

Newberry, B.H., D.R. Baldwin, J.E. Madden, and T.J. Gerstenberger 1987 Stress and disease: an assessment. Pp. 123-151 in J.H. Humphrey, ed., Human Stress: Current Selected Research, Vol. 2. New York: Ames Press.

Novaco, R.W. 1988 Stress Reduction Programs. Background paper prepared for the Committee on Techniques for the Enhancement of Human Performance, National Research Council. Washington, D.C.: National Academy Press. Available through the National Academy Press's Publication-on-Demand Program.

Park, W.-W. 1990 A review of research on groupthink. Journal of Behavioral Decision Making 3:229-245.

Parkes, K.R. 1986 Coping in stressful episodes: the role of individual differences, environmental factors, and situational characteristics. Journal of Personality and Social Psychology 51:1277-1292.

Paterson, R.J., and R. Neufeld 1987 Clear danger: situational determinants of the appraisal of threat. Psychological Bulletin 101:404-416.

Payne, J.W., J.R. Bettman, and E.J. Johnson 1988 Adaptive strategy selection in decision making. Journal of Experimental Psychology: Learning, Memory, and Cognition 14:534-552.

Reason, J. 1988 Stress and cognitive failures. Pp. 405-421 in S. Fisher and J. Reason, eds., Handbook of Life Stress, Cognition and Health. New York: Wiley. 1990 Human Error. New York: Cambridge University Press.

Ritov, I., and J. Baron 1990 Reluctance to vaccinate: omission bias and ambiguity. Journal of Behavioral Decision Making 3:263-277.

Rogers, K.J.S., M.J. Smith, and P.C. Sainfort 1990 Electronic performance monitoring, job design and psychological stress. Pp. 854-858 in Proceedings of the Human Factors Society 34th Annual Meeting. Santa Monica, Calif.: Human Factors Society.

Rothstein, H.G. 1986 The effects of time pressure on judgment in multiple cue probability learning. Organizational Behavior and Human Decision Processes 37:83-92.

Schneider, W., and R.M. Shiffrin 1977 Controlled and automatic human information processing: I. detection, search, and attention. Psychological Review 84:1-66.

Selye, H. 1956 The Stress of Life. New York: McGraw-Hill. 1983 The stress concept: past, present, and future. Pp. 1-20 in C.L. Cooper, ed., Stress Research . New York: Wiley.

Shanteau, J., and G.A. Dino 1993 Environmental stressor effects on creativity and decision making. Pp. 293-308 in O. Svenson and A.J. Maule, eds., Time Pressure and Stress in Human Judgment and Decision Making. New York: Plenum.

Shiffrin, R.M., and W. Schneider 1977 Controlled and automatic human information processing: II. perceptual learning, automatic attending, and a general theory. Psychological Review 84:127-190.

Siconols, M. 1992 Individual investors' holdings of U.S. stocks fall below 50% of total market for the first time. The Wall Street Journal 13(November):C1.

Smith, R.C. 1985 Stress, anxiety, and the air traffic control specialist. Pp. 337-358 in I.G. Sarason and C.D. Spielberger, eds., Stress and Anxiety , Vol. 9. New York: Hemisphere Publishing.

Time 1983 June 6 issue.

U.S. Army 1992a Fratricide: reducing self-inflicted losses. CALL Newsletter 92(April):4. Fort Leavenworth, Kans.: U.S. Army Combined Arms Command. 1992b Fratricide risk assessment for company leadership. CALL Handbook 92(March):3. Fort Leavenworth, Kans.: U.S. Army Combined Arms Command.

Voss, H.-G. 1977 The effect of experimentally induced activation on creativity. Journal of Psychology 96:3-9.

Wickens, C.D. 1992 Engineering Psychology and Human Performance, 2nd ed. New York: Harper Collins.

Wickens, C.D., A. Stokes, B. Barnett, and F. Hyman 1989 The Effects of Stress on Pilot Judgment in a MIDIS Simulator (U). Armstrong Aerospace Medical Research Laboratory Report No. AAMRL-TR-88-057. Wright-Patterson Air Force Base, Ohio: Army Aerospace Medical Research Laboratory.

Williams, H.P., M. Tham, and C.D. Wickens 1992 Resource Management and Geographic Disorientation in Aviation Incidents: A Review of the ASRS Data Base. Aviation Research Laboratory Report No. ARL-92-3/NASA-92-2. Savoy: University of Illinois.

Woods, D.D. 1993 Process-tracing methods for the study of cognition outside of the experimental psychology laboratory. Pp. 228-251 in G.A. Klein, J. Orasunu, R. Calderwood, and C.E. Zsambok, eds., Decision Making in Action: Models and Methods. Norwood, N.J.: Ablex.

Yates, J.F. 1990 Judgment and Decision Making. Englewood Cliffs, N.J.: Prentice Hall.

Zajonc, R.B. 1968 Social facilitation. Science 149:269-274.

Zakay, D., and S. Wooler 1984 Time pressure, training and decision effectiveness. Ergonomics 27:273-284.