National Academies Press: OpenBook
« Previous: 2 Analogous Systems
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 54
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 55
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 56
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 57
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 58
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 59
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 60
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 61
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 62
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 63
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 64
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 65
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 66
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 67
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 68
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 69
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 70
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 71
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 72
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 73
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 74
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 75
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 76
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 77
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 78
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 79
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 80
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 81
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 82
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 83
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 84
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 85
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 86
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 87
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 88
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 89
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 90
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 91
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 92
Suggested Citation:"3 Workload Factors." National Research Council. 1993. Workload Transition: Implications for Individual and Team Performance. Washington, DC: The National Academies Press. doi: 10.17226/2045.
×
Page 93

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Workload Factors A characteristic of most post-transition periods is a large number of task demands often imposed with very severe time constraints. These tasks are often characterized by the description of high workload. The term workload has intuitive meaning for most people; everyone has experienced periods of high or low workload in their daily life in response to different situations. Psychologists have invoked the concept in theories of attention and performance. Aircraft designers, manufacturers, operators, and regula- tory agencies have identified operator workload as a critical factor in sys- tem effectiveness. However, the word workload did not appear in many dictionaries until the 1970s; and operational definitions proposed by psy- chologists and engineers continue to disagree about its sourcets), mechanisms, consequenceLs), and measurement. Furthermore, although workload and performance are clearly related, it has proven to be a much more complex relationship than originally thought. WORKLOAD CHARACTERISTICS Sources Over the past 20 years, workload has been equated with: (1) imposed task demands if the difficulty, number, rate, or complexity of the demands imposed on an operator are increased, workload is assumed to increase; (2) the level of performance an operator is able to achieve if errors increase or control precision degrades, workload is assumed to increase; (3) the 54

WORKLOAD FACTORS 55 mental and physical effort an operator exerts-workload reflects an operator's response to a task, rather than task demands directly; and (4) an operator's perceptions if an operator feels effortful and loaded, then workload has, in fact, increased even though task demands or performance have not changed. Most contemporary definitions assume that workload emerges from the in- teraction between a specific operator and the assigned task (Gopher and Donchin, 1986; Hart, 1986; Hart and Wickens, 19901. Consequences The consequences of optimal or suboptimal levels of workload depend on the structure of a specific task, the environment in which it is performed, and operator characteristics. As task difficulty increases, performance of- ten, but not always, degrades; response times and errors increase for dis- crete tasks, control variability and error increase for tracking tasks, and fewer tasks are completed within an interval of time. The workload im- posed by one task may interfere with the performance of other concurrent activities. The subjective experience of excessively low or high workload may prompt operators to adopt different task-performance strategies. Pro- longed periods of high workload may result in operator fatigue. An in- crease in psychological stress that often accompanies high workload may result in elevated heart rate. Measures To some extent, these apparent complexities have been caused by (1) the practice of using the same term (workload) to refer to the demands imposed on an operator; the effort exerted to accomplish those demands; and the physiological, subjective, or performance consequences of an operator's actions and (2) the naive assumption that different measures of workload index the same entity. In fact, different measures are sensitive to different aspects of workload and are appropriate for answering different questions. Thus, the results of most workload studies are interpreted far too broadly, given the complexity of the phenomena and the limited range of factors to which each type of measure is sensitive. The problem is compounded by the absence of a generally accepted definition, standardized procedures and units of measurement, or an absolute standard against which to compare a particular task or candidate measure. Relationship Between Workload and Performance Designers, manufacturers, and operators of complex systems have been more interested in the association between workload and performance than

56 WORKLOAD TRANSITION in theoretical issues. They assume that human performance is most reliable under moderate workload that does not change suddenly or unpredictably (Kantowitz and Casper, 19884. When workload is too high, errors arise from an operator's inability to cope with critical task demands. When workload is too low, errors may arise from loss of vigilance and boredom (see Chapter 6~. However, the actual relationship between the effort an operator invests in a task and the performance he or she is able to achieve is much more complex. O'Donnell and Eggemeier (1986) suggested that the relationship be- tween workload and performance changes as overall task difficulty is in- creased. For relatively easy tasks, operators can maintain consistent perfor- mance by exerting additional effort if task demands are increased. For moderately difficult tasks, they cannot maintain consistent perfo~-~ance even if they exert additional effort. For extremely difficult tasks, operators do not have the capacity to exert additional effort in response to further in- creases in task demands, so performance decrements no longer reflect a change in workload. Navon and Gopher (1979) suggested that this relationship depends on the structure of the task, differentiating between situations in which addi- tional resource investment would (resource-limited) or would not (data- limited) result in improved performance. For data-limited tasks, additional effort cannot improve performance. If sufficient information is available, the task can be performed. If it is not, the task cannot be performed ad- equately. For resource-limited tasks, however, additional effort can result in improved performance. Hart (1989) and Hart and Wickens (1990) discussed the importance of operator strategies in determining the relationship between workload and performance. People may not try to achieve perfect performance or accom- plish tasks immediately. Rather, they manage their attention and effort, rescheduling, deferring, or shedding less important tasks to achieve accept- able performance and maintain a reasonable level of workload for the dura- tion of the task (see Chapter 9 for some discussion of how effectively this scheduling is accomplished). In a recent interview, Broadbent (1990:7' stated that humans are "dynamic, self-regulating systems, with each func- tion monitored and modified by others." Thus, to optimize their interaction with a complex system, operator beliefs, values, and intentions must be considered. Because it is clear that humans have interests beyond optimiz- ing traditional indices of system performance (Rouse, 1979), simple linear models cannot provide an accurate description of their behavior. Despite the apparent confusion about what workload is (one might fairly say that anything can cause workload) and what its effects might be, there are some factors that stand out as being particularly salient. The duties assigned to the crews of tanks, helicopters, jets, nuclear power plants, and so on impose a set of requirements that vary in magnitude alla composition

WORKLOAD FACTORS 57 from one moment to the next, from one type of activity or mission phase to another, and from one crew position to another. The operational require- ments of the tank crew are composed of a variety of behaviors that have been studied in isolation (in laboratory research) or in different combina- tions (in simulation and field studies) to determine the specific task-related, equipment-related, environmental, and operator-related factors that influ- ence workload and performance. From this research, general principles may be drawn to estimate the effects of performing specific activities on the operators of a particular system performing a mission. The following pages contain a review of data regarding the primary factors that drive workload, showing how these drivers are relevant to a variety of transition teams. WORKLOAD DRIVERS: REVIEW OF RESEARCH The intrinsic difficulty of the activities that an operator must perform establishes the target or nominal level of workload. The difficulty of a particular task may be influenced by any one or several of the following factors: (1) the goals and performance criteria set for a particular task; (2) the structure of the task; (3) the quality, format, and modality in which information is presented; (4) the cognitive processing required; and (5) the characteristics of the response devices. Although the fundamental source of workload is literally the "work" that is "loaded" on an operator, the behav- ior and workload experiences of a particular operator may be influenced by other factors as well. For example, fatigue, stress, training, crew coordina- tion, and environmental stressors (e.g., heat, cold, vibration, noise, and dan- ger) may have a significant impact on operator workload. Although the influence of these factors is obvious in operational situations, little research has been performed to establish their relationship with operator workload, and they are ignored in most workload theories. The relationship between these factors and team performance are reviewed in other chapters. Given the limitations of human memory, vision, physical strength, and so forth, some tasks may stretch or even exceed an operator's capacities; other tasks impose so few demands that they may be performed concur- rently with other tasks. Thousands of experiments have been conducted in which a variety of task-difficulty manipulations were imposed to examine different aspects of the human information processing system and to define and quantify the limits of human capabilities. The following section sum- marizes the results of a representative sampling of this work. Task Structure The way in which a task or combination of tasks is organized, the rate at which information is presented or error signals change, the length of time

58 WORKLOAD TRANSITION the task must be performed, and the levels of speed and accuracy that the operator must try to achieve have a significant impact on the workload imposed during its performance. A significant number of the experiments conducted to analyze human operators' responses to different levels of workload have included some variation in the rate of presentation for discrete tasks or disturbance bandwidth for control tasks. Thus, a considerable amount is known about these factors. Relatively less information is available about the effects of speed-accuracy tradeoffs, task schedules, and task duration. Performance Criteria and Strategies The difficulty of almost any task can be altered by a requirement for additional speed or accuracy. As criteria for acceptable performance be- come more stringent, the workload associated with attaining adequate per- formance increases (Yeh and Wickens, 1988~. People may adopt either externally imposed performance criteria or personal criteria (which may or may not be more stringent). Workload and performance are influenced by the objective consequences of failing to meet task requirements as well. People are more likely to try to meet performance standards if their job or personal safety are on the line, than if the consequence of poor performance is simply a "bad score." In addition, operators may act less conservatively, take more risks, and try new techniques when there are no dire conse- quences of failure. In general, manual controllers of dynamic systems such as the tank driver try to follow a particular course or path, while minimizing the effects of external disturbances. Error will increase if control inputs are too little, too late, or in the wrong direction. If operators overcontrol (i.e., make control inputs that are faster or larger than necessary), they create additional workload for themselves (i.e., they must compensate for the errors that they generate) and run the risk of destabilizing the system. Thus, although mini- mizing error is the goal of most control activities, smoothness and stability may be equally important. Similarly, in discrete control tasks like target acquisition, performance strategies can be described by a tradeoff between the speed and accuracy of the movements. Faster movements are made with less accuracy, and more precise movements are made more slowly. This tradeoff is described by a mathematical model known as Fitts's Law. This relationship has proven to be extremely robust (Keele, 1986) for a wide range of target types, widths, and distances, limbs (e.g., fingers, arms), system dynamics (e.g., displace- ment of the joystick controls cursor position or velocity), control devices (e.g., computer mouse, joystick, rotary knob), and displays (e.g., computer screen, direct view of nearby or distant target). More recent research has demonstrated that Fitts's Law also holds for targets that vary dynamically in

WORKLOAD FACTORS 59 size (Johnson and Hart, 1987) or position (Jagacinski et al., 1980~. A similar relationship has been found between the movement difficulty and subjective workload (Hart et al., 1984b; Johnson and Hart, 1987; Mosier and Hart, 1986~. Instructions to maximize either speed or accuracy influence the workload and performance associated with discrete responses, target acquisition, and continuous control. If operators are instructed or choose to maximize preci- sion, the cost may be an increase in time for task completion. Conversely, if they maximize speed, then accuracy may be reduced. Depending on the situation, either speed or accuracy may be more important. Thus, the over- all quality of performance depends on the operators' correct assessment of which factor is most critical. For example, correctly identifying a target (as friend or foe) is more important than firing quickly (and running the risk of hitting a friend), even though verifying the identity of a target may take longer than making a snap judgment. If the position of a helicopter or tank is not yet known to an enemy, making the first shot count (e.g., accuracy) is extremely important, as firing a weapon will reveal its position. Alterna- tively, if a tank is being fired on or the engine of a helicopter fails, a rapid response is essential. Smooth control and on-time arrival may be more important to an airline pilot than keeping the needles precisely centered on the intended flight path, while extreme precision is required to accomplish inflight refueling. Different components of a complex task may have different functional priority with respect to the overall goals of the task or temporal priority (if they are not completed by a deadline, they can no longer be performed at all). Operators may adopt different resource allocation and scheduling strategies to satisfy external instructions or personal goals. Within limits, people are able to maintain a particular level of performance on one task (at the ex- pense of another) or to share attention equally or in graded amounts be- tween two tasks (Gopher et al., 1982~. They may adopt a fixed policy for resource allocation or dynamically modulate it in response to changes in task demands or priority over time. People are better able to dynamically allocate the same resource across tasks in response to priority instructions (e.g., devote more processing resources to a tracking task than to a concur- rent running memory task) than to dynamically allocate graded amounts of different resources (e.g., devote more visual resources to one task than auditory resources to another). Shifting task priorities and resource alloca- tion policies in response to transient changes in task demands is particularly difficult (Tsang and Wickens, 1988) and may inhibit an effective response to workload transitions. In practice, humans tend to maintain a particular strategy of resource allocation, even though the situation has changed. To achieve required levels of performance on competing tasks, subjects try to perform tasks simultaneously, or allow high-priority tasks to preempt

60 WORKLOAD TRANSITION their attention whenever their performance begins to degrade. In opera- tional situations, task prioritization may be formalized, based on safety or mission-related concerns. For example, all pilots are taught to aviate, navi- gate, and communicate, in that order of importance. However, even low- priority tasks can assume temporal priority (i.e., verbal communications do not wait for an operator to get around to them) because they grab attention (they are loud, bright, unexpected, etc.) or because they can be completed quickly (thereby getting them out of the way). An operator's ability to respond to changing priorities is influenced by the total demands of concurrent tasks, their resource requirements, instruc- tions, feedback, and training (Tsang and Wickens, 19881. Although little research has been performed to relate priority manipulations and dynamic resource allocation to workload, it is reasonable to assume that workload is likely to increase when performance strategies must be changed in response to a shift in task priority. Furthermore, if operators select an inappropriate strategy (e.g., focusing more resources than required on one task and too little on another), resources are wasted with a resulting cost in performance, workload, or both. It appears that people can learn more efficient methods of responding to task-specific, priority manipulations through training (Gopher et al., 1989~. Furthermore, more generic skills developed in one task (i.e., a video game) can transfer to another (i.e., flight), thereby facilitating performance (Go- pher et al., 19881. Task Schedule The way operators organize their time and resources to perform com- plex tasks has a significant impact on the workload experienced and perfor- mance achieved. To some extent, task scheduling depends on external fac- tors (e.g., instructions, procedures, information availability, deadlines, the length of time a task can be safely ignored, etc). In laboratory research, tasks that are to be performed sequentially are generally presented sequen- tially and those that are to be performed concurrently are presented concur- rently, to elicit a particular response strategy. In realistically complex situ- ations, however, operators may have greater flexibility in how they perform multiple tasks. They may choose to perform different task combinations either simultaneously, alternatively, or sequentially, with different workload and performance costs. Sequential performance may decrease workload and result in a better quality of performance. However, this strategy may delay the completion of some task components. Concurrent performance may increase workload and result in a poorer quality of performance, although the time to complete all of the tasks may be less. When tasks are performed simultaneously,

WORKLOAD FACTORS 61 workload and performance depend on the cognitive and physical resources required. If two tasks require the same resources, and if their demands exceed available capacity, performance on one or both of them will suffer. When operators alternate between several tasks (or perform them sequen- tially), resource competition is no longer an issue, although switching atten- tion may require additional time and effort (see Chapter 9~. In addition, switching from one stimulus-response modality to another takes more time and effort than switching between tasks that require the same stimulus- response modalities. For example, visual-spatial target acquisitions take longer and impose higher workload when the correct target is identified auditorially, rather than visually, or represented verbally rather than spa- tially (Hart et al., 1986~. If operators are allowed some scheduling flexibility, they may adopt different strategies depending on the characteristics of a particular task com- bination. For example, King et al. (1989) found that subjects faced with performing a three-axis tracking task while also completing a series of discrete tasks adopted both time-sharing and switching strategies. Discrete tasks that could be completed quickly preempted the tracking task. Sub- jects were able to maintain single-task performance levels on the discrete tasks without seriously disrupting control performance because they required little time away from tracking. Subjects also alternated between the control task and discrete tasks that took several seconds to complete. Response times increased significantly over single-task levels, and subjective workload was high, possibly reflecting the added cost of switching back and forth. Moray and Liao (1988) found that when the presentation rate of four con- current tasks was increased until there was insufficient time to service all of the tasks, subjects simply stopped performing some of the tasks to protect performance on the others. When performing familiar tasks, in which variations in task demands are relatively predictable, completing some tasks ahead of schedule and developing contingency plans during periods of low workload can reduce workload and improve performance during later periods of high workload. For example, Pepitone et al. (1988) found that contingency planning signifi- cantly improved the quality of later decisions made under time stress and resulted in safer flights. Furthermore, if the time required to complete specific tasks and the time remaining in which to complete them are known, then operators can develop more efficient task-performance schedules. When events unfold as planned and well-rehearsed sequences of actions can be relied on, cognitive demands are lower, performance better, and workload less than when unexpected events occur. Thus, developing a range of contingency plans for potential events is a valuable workload man- agement tool. Planning ahead requires time and effort, however. For ex- ample, Vienneau and Gozzo (1987) found that helicopter pilot workload

62 WORKLOAD TRANSITION was higher during the planning stage immediately preceding the execution of a maneuver than during the maneuver itself. Thus, predicted load levels (based on physical activities) lagged the actual load levels by one event. Hart and Hauser (1987) obtained similar results with fixed-wing pilots. Premission workload levels were higher than those of all but the most de- manding flight segments. In laboratory research, random intertrial intervals and lack of preview about upcoming events generally preclude anything but a reactive strategy. However, the little research that has been performed on this topic suggests that people are able to develop more efficient strategies if given the oppor- tunity. For example, Tulga and Sheridan (1980) examined decision makers' performance in a dynamic, multitask simulation. The number, durations, deadlines, and payoff of tasks and intertask arrival times were manipulated. With moderately complex schedules, operators adopted relatively optimal scheduling strategies. However, they reverted to a reactive strategy when the situation became too complex. Using a similar paradigm, Hart et al. (1984a) found that proficient subjects adopted a time-sharing strategy, rather than a sequential strategy, and actively controlled the flow of task elements (e.g., requesting tasks ahead of schedule during periods of low workload, deferring some tasks during periods of high workload, and shedding tasks they could not complete), thus completing twice as many tasks in the same interval of time as low-scoring subjects, who were more reactive and per- formed task elements sequentially, even though there was sufficient time to alternate between tasks. In operational situations, team members may request assistance from each other as a strategy for coping with excessive task demands. Although the effects of team workload management strategies are not well known, it seems reasonable to assume that the appropriate management of human and system resources can result in adequate performance and acceptable workload, even in the face of relatively high task demands. A team commander must monitor the workload of team members and shift responsibilities to avoid an uneven distribution of workload across people and time. However, when a crew adopts a reactive strategy or is forced into one by externally paced and unpredictable task demands, performance is more likely to suffer and workload will be higher. Rate of Presentation Almost any task within an operator's capabilities can be performed correctly and with acceptable workload if sufficient time is available. Con- versely, if sufficient time is not provided to perform a single task, or if the interval between task elements is reduced below some critical value, then task completion is not possible, no matter how much effort is invested. In

WORKLOAD FACTORS 63 between, where internal or external constraints impose some time pressure but task performance is still feasible, the usual finding is that subjective workload increases as time pressure increases. However, people may adopt different coping strategies to deal with an increase in time pressure (e.g., trade accuracy for speed, shed or defer some tasks, reduce their perfor- mance criteria), thereby reducing the effects of time pressure on their workload. There are structural limitations in the speed with which humans can perceive, process, and respond to successive inputs. For example, percep- tual events that occur within roughly 100 milliseconds of each other are combined into a single perceived event (Card et al., 1986~. If one stimulus is presented before a previous one has been fully processed, responses to the second will be delayed, presumably because the central processor can operate on only one task at a time (Welford, 19671. Card et al. (1986) estimated the duration of the entire perception-processing-action feedback loop to take between 200 and 500 milliseconds. Thus, since the discrete micromovements that combine to create continuous control activities re- quire only about 70 milliseconds (Keele, 1986), a series of motor move- ments can be executed open-loop between corrections guided by visual feedback. For discrete responses, the lower limit of reaction time is in the range of 70- 100 milliseconds. However, the minimum time required to complete any particular keystroke or button press is determined by the duration of the perceptual and cognitive processes required to select the correct response. For example, Card et al. (1986), in their summary of research, identified the minimum processing times associated with: (1) number of potential re- sponses (response times increase by 90-150 milliseconds for each additional alternative), (2) comparisons with remembered items in short-term memory (30-70 milliseconds for each additional item, depending on their complex- ity), (3) meaningfulness (a range of 158 to 500 milliseconds per typing keystroke for text, random words, and random letters, with text being the fastest and random letters the slowest), (4) matching stimulus input to an internal representation (310, 380, and 450 milliseconds, respectively, for physical, name, and class matches), and so on. Obviously, the minimum time required to perform the complex activi- ties typical of the real world is longer and more difficult to predict. However, the same principles hold. If operators are not allowed sufficient time to encode and process information, select a response, and execute the response, then performance quality will suffer and higher workload will be experienced. In continuous control tasks, such as that performed by the tank driver, the operator's goal is to minimize the time-averaged difference between a target (e.g., a nominal flight path, a driving lane, a point or line on a screen) and the output of a dynamic system. This goal is accomplished by manipu- lating available control mechanisms (e.g., aircraft control inputs, steering commands, or joystick inputs). The difficulty of this task is directly influ

64 WORKLOAD TRANSITION enced by the predictability, frequency, and amplitude of the disturbances (e.g., winds) for which the operator is trying to compensate. People may be able to compensate perfectly for highly predictable, low-frequency distur- bances; error and workload generally increase at higher frequencies (see, for example, Moray and Liao, 1988~. Random or quasi-random signals composed of many frequencies are less predictable and therefore impose even higher workload. As more frequencies are combined to create a com- plex signal, it becomes more difficult for an operator to track. An increase in bandwidth (the upper limit of the frequencies represented in a complex signal), such as that caused by driving an unpredictable course at faster speeds is usually associated with an increase in workload as well (see, for example, Hauser et al., 1983; King et al., 1989~; operators must enter cor- rections and monitor the visual display more often with high-bandwidth disturbances. However, people may be less sensitive to the effects of band- width on workload than they are to other manipulations of tracking task difficulty (Vidulich and Wickens, 1984, 1986~. There is evidence from scores of dual-task experiments that low-bandwidth tracking tasks (e.g., less that 0.5 Liz) allow the operator sufficient time to simultaneously perform many types of concurrent tasks. However, it is difficult for them to main- tain the same level of performance on high-bandwidth tasks (e.g., greater than 1.0 Hzj while also performing other activities. Maintaining acceptable performance when a signal is very unpredictable simply demands too much time and attention. The relationship between rate of presentation (or response) and workload or performance may be U-shaped. For extremely slow or infrequent tasks, loss of vigilance has been associated with slower response time and higher subjective workload; boredom is unpleasant for many people (Hancock and Warm, 1989~. A moderate increase in presentation rate may increase arousal and result in faster response times, but at the cost of an increase in workload (see, for example, Moray and Liao, 19883. Beyond some point, further in- creases in presentation rate generally result in an increase in subjective workload, errors, or delayed responses as operators attempt to share their attention be- tween temporally overlapping activities. For example, Pepitone et al. (1987) found that pilot workload was significantly positively correlated with the rate at which flight-related tasks were imposed while pilots flew a simulator. As presentation rate was increased from once every 2.4 minutes to once every 0.8 minutes, subjective workload increased significantly. For complex tasks, the pattern of performance decrements that occurs when there is insufficient time to monitor, process, and respond to all task components depends on the strategy an operator adopts. For example, Mo- ray and Liao (1988) found that increased task frequency had different ef- fects on the performance of each of four concurrent tasks. For choice reaction time and arithmetic tasks, response time became faster but accu

WORKLOAD FACTORS 65 racy suffered as task pacing increased. In addition, tracking error increased and target identification responses were delayed. In this experiment? there was no catastrophic collapse in performance, but rather a graceful degrada- tion controlled by the operator's tactical allocation of attention; the operator appeared to give high priority to the choice response task and low priority to the relatively infrequent monitoring task in order to manage workload. Complexity of Task Demands As the number of alternative choices among which an operator must select increases, the time required to respond correctly generally increases. Hick (1952) and Hyman (1953) postulated the Hick-Hyman Law, which asserts that response latency increases linearly as the logarithm of the num- ber of alternatives is increased. Later research demonstrated a reliable relationship between the number of alternatives and workload. For ex- ample, Kantowitz et al. (1984) found that performance was worse and workload was higher for a four-choice task than for a two-choice task in the context of simulated flight. Furthermore, performance on tasks with many alterna- tives is more likely to be disrupted by concurrent activities than is perfor- mance on a single-alternative response task. As the complexity of a task or combination of tasks performed concur- rently increases, workload often increases as well. For example, Bittner et al. (1988) found that the workload of mobile air defense system crews was significantly higher when attacked by two helicopters than by one in a field test. lIurst and Rose (1978) found that air traffic controllers reported higher workload as the number of aircraft under their control increased. However, it is the cognitive complexity of a task that influences workload, more than the absolute number of task elements (Kantowitz and Casper, 19881. A1- though Hart and Hauser (1987) found that communications frequency was significantly correlated with the workload of test pilots, more detailed analyses of pilot communications suggested that other factors have an even more significant impact on workload: delayed, unexpected, or very complex messages or clearances that result in a change of plans impose higher workload than do routine messages conveying expected information (Acton et al., 1983; Hart and Bortolussi, 1984~. Aretz (1990) found that map complexity had a more powerful effect on the cognitive operations required for naviga- tion than any other variable; each additional landmark added to the map increased response time by 450 milliseconds. Variability of Task Demands People develop expectations about how much effort a specific task should require. If a particular instance is more difficult than expected, workload

66 WORKLOAD TRANSITION will seem even higher than it actually is. Furthermore, people may select a strategy that is inappropriate for the current situation, based on expectations ~ . . from prior experiences. Operator workload may vary dramatically over time within a task. As load levels increase or decrease, the rate or intensity of effort exerted dur- ing the immediately preceding interval may influence subsequent behavior. Thus the task performance strategies operators adopt, and the level of per- formance they achieve, cannot be predicted from the current load level alone (Matthews, 1986~. Instead, when task demands change cyclicly or randomly, people may persist with a previously successful strategy (Poulton, 1982) or fail to anticipate or recognize a change in task demands (summing and Croft, 1973~. Thus, they maintain an inappropriately high response rate (when task demands have shifted from high to low) or an inappropriately low response rate (when task demands have shifted from low to high). In the first case, persistence of the previous strategy may benefit performance but result in unnecessarily high workload; in the second case, performance will degrade. Rapid changes in demand level may result in a sudden change in arousal level and require a number of resources to be recruited for task perfor- mance. This shift takes time and requires effort. For example, Thornton (1985) demonstrated that an unexpected peak of workload introduced in a relatively low workload task, elevated subjective workload for the whole segment, particularly if it occurred late in the interval. If task difficulty was changed more gradually, however, performance was not disrupted and rated workload more accurately reflected the average demands of each seg- ment in a series of target acquisition tasks (Staveland et al., 1986~. Hart and Bortolussi (1984) examined pilots' opinions about the effects of sudden changes in task demands. They found that almost any additional task, deviation in flight plan, or system failure resulted in increased workload and degraded performance, particularly when changes occurred during phases of flight that already imposed high workload. On the average, routine events or activities increased workload by 6 percent, system failures in- creased workload by 30 percent, and pilot errors increased workload by 16 percent. The effect of pilot error on subsequent performance and workload was particularly interesting. Through experience, pilots had come to expect that additional errors would follow those already committed and that workload would be elevated when resolving an error (i.e., they must recognize the occurrence of an error and then develop and execute a plan for resolving it). Errors may force an operator out of routine, low workload, patterns of behavior and, in fact, completely change the character of the task until the error is resolved. Significant increases in stress were associated with opera- tor errors and system failures, even when the effort required to deal with the l

WORKLOAD FACTORS 67 problem was low. In contrast, other events (e.g., communications) increased workload but did not change stress. Task Duration The relationship between task demands, effort, and performance is moderated by task duration. People may be able and willing to exert considerable effort or accept inactivity and boredom for brief intervals, but not for very long. Particularly in familiar and predictable situations, experienced opera- tors will pace themselves, working at a rate and effort level that they can sustain for the expected duration of the task. If they do not pace themselves appropriately (either because they are inexperienced or the situation is dif- ferent than expected), then performance is likely to suffer as the mission progresses. Good leadership ensures that members of a team can maintain performance throughout an entire mission by setting an appropriate pace. When performing prolonged tasks, particularly those that are repetitive and monotonous or present little task-relevant information, people become bored. When this occurs, performance becomes less efficient (e.g., opera- tors make errors, miss relevant signals, respond less frequently or more slowly, or change their decision criteria) (Hockey, 1986~. Fatigue may arise from excessive demands imposed on an operator for a prolonged period. Physical and mental fatigue may occur as a direct result of the effort ex- erted to perform a task, while emotional fatigue may reflect prolonged ex- posure to the stressful environment in which a task is performed (Rotondo, 1978~. Thus, when difficult or stressful activities must be performed for a prolonged period, operators' capacities may be diminished and their perfor- mance will suffer unless they exert additional effort. These effects are discussed in more detail in Chapters 5 and 6. Although some of the subjec- tive symptoms of boredom are similar to those of fatigue, the effects of boredom are more easily reversed by a change in the situation. Within limits, however, task duration alone does not appear to have a consistent influence on workload or performance. Research has shown that difficulty, stimulus-response compatibility, resource competition, presenta- tion rate, and so on have a more significant effect. For example, in most simulation and flight research, workload ratings are obtained for each seg- ment of flight. Even though takeoff and landing segments last only a few minutes and cruise segments last considerably longer, pilot workload rat- ings reflect the difficulty rather than the duration of a segment (Bortolussi et al., 1987; Hart and Hauser, 1987~. Pepitone et al. (1987) developed flight scenarios in which segment length and within-segment difficulty were not confounded. Here, too, workload covaried with the number and frequency of tasks performed within a segment, not with flight segment duration. In

68 WORKLOAD TRANSITION fact, workload was lower during longer segments than during shorter seg- ments. Similarly, Hart and Hauser (1987) found that overall workload and subjective fatigue were unrelated to the duration of flights performed at night in the Kuiper Airborne Observatory of the National Aeronautics and Space Administration. Task Requirements and Procedures i It should be clear from the preceding discussion that the introduction of automation must be preceded by a careful analysis of task requirements and human capabilities. Such analysis is also critical for the distribution of tasks across people within a team. The design questions here concern who should perform which tasks to maintain an adequate balance of information processing load across operators and machines, yet to operate within the constraints of the command hierarchy within the system. Should tasks be redistributed to other operators because of overload? Or should they be redistributed to computers? which are lighter in weight and, for many tasks, more reliable? How should the operator monitor the state of an automati- cally controlled system? If tasks are shifted or shared between operators, is the reduction in task load on the two operators greater than the increase in communications load necessary to coordinate between them? The potential to address these sorts of complex multielement design issues is provided by interactive computer simulation design tools, which incorporate models of task scheduling, crew interaction, anthropometry, dis- play layout, and operator time-sharing (Elkind et al., 1990~. Two parallel approaches potentially have much to offer at this level: the CREWCUT project sponsored by the U.S. Army Human Engineering Laboratory (Hahler et al., 1991; Lockett et al., 1990; Prevost and Banda, 1991; Wickens, 1992a) focuses directly on developing a simulation model for tank crew operators; and the MIDAS (Man-Machine Integration, Design, and Analysis System) project developed at NASA Ames Research Center (Bamba et al., 1991), although it originally focused on rotorcraft, is designed for a broader range of multicrew systems. The research effort on modeling distributed decision making carried out by Kleinman and his colleagues, described in Chapter 10, would appear to have considerable potential to predict the implications of different organizational structures on team performance. Finally, the issue of task requirements is closely related to the proce- dures that are either mandated or, in some cases, relieved by operating procedures. The Federal Aviation Administration, for example, has adopted a host of operating procedures intended to improve air safety, although adherence to some of these will not necessarily decrease workload and may increase it. Adopting policies mandating adherence to a sleep schedule (see Chapter 5) certainly falls in this category. New procedures and regulations,

WORKLOAD FACTORS 69 however, should be introduced with caution and careful analysis of their potentially less obvious side effects. Sometimes the costs (and burden on the operator) to implement may far outweigh the benefit, particularly if the regulation was introduced with the intention of "locking out" a very infre- quent operator error, whose cause could not be established with certainty in the first place. r Input Variables Operators require information to perform most tasks. They may obtain necessary information from memory; however, this section addresses infor- mation gained from external sources through an operator's eyes and ears while the task is being performed. The operators of complex systems are bombarded by visual and auditory inputs. Their perceptual systems trans- late the information detected by their eyes and ears into internal representa- tions that may be subject to deeper analysis. Because so much information may be presented at the same time, operators must divide their attention among these different sources, seeking, interpreting, and integrating rel- evant information. There is no question that crew workload is influenced by the number and complexity of the information sources that must be monitored and the quality of the information they provide. However, the influence of purely perceptual factors on workload has been examined in less depth than have information processing, memory, and response mecha- nisms. The workload associated with perceiving an auditory or visual signal depends on a number of factors, such as intensity (i.e., whether or not it is above threshold), signal/noise ratio (i.e., the similarity of the signal to the background in which it is presented), rate of change (i.e., whether a change is fast enough or large enough to be noticeable), distinctiveness, frequency, and familiarity, among other factors. If it is difficult to identify a signal, workload is likely to increase. Information From Visual Displays Analog and digital instruments provide information about the state of the system and, in many vehicles, additional information is provided about the current location, orientation, speed, altitude, depth, and heading of the vehicle and the environment through which it is moving. The display for- mat may be analog or digital, abstract or representational, and one- or mul- tidimensional. The reference coordinate systems (e.g., earth-centered, ve- hicle-centered) differ across displays. Because the quality, format, and content of displays vary, the perceptual demands their use imposes on an operator also vary. And, because there may be many information sources that an

70 WORKLOAD TRANSITION operator must monitor, formal procedures may be established to ensure that visual scan patterns sample different displays often enough and that com- peting auditory messages are prioritized. Electronic displays are beginning to replace analog gauges in most air- craft, power plants, automobiles, and other complex systems. All too often, digital readouts have replaced analog. Whereas pilots of earlier aircraft simply had to check the general orientation of a number of needles to assure themselves that system state was normal, digital readouts must be read individually, present unnecessarily precise information for most uses, and do not provide readily interpreted trend information. Since analog displays can be checked more quickly and with less workload than digital readouts (Hanson et al., 1981), the influence of this design decision is obvious. Furthermore, an alphanumeric display (which is represented in memory ver- bally) must be transformed- into a spatial representation to be integrated into the pilot's internal model of the environment, vehicle orientation, and state. Display design should foster an appropriate distribution of attention, given the relative importance of different display elements. If a display is poorly designed, it may capture the operator's attention for an inordinate amount of time to extract relevant data. Thus, other important sources of information inside or outside the cockpit may be ignored. For example, Harris et al. (1982) found that different analog display formats and place- ments required significantly different amounts of visual attention and cre- ated different levels of workload. They found that a bar graph type of vertical speed indicator located in a more logical position on the panel could be interpreted more quickly, thus reducing pilots' mental workload. The more appropriate format and placement required shorter visual dwell times. Interface design affects both workload and the cognitive strategies adopted (Woods, 1989~. For example, the design of display elements and their organization can force even a highly skilled user from automatic perceptual processes to effortful cognitive processes, or vice versa. Particularly for displays with which complex, automated systems are monitored, the format should promote the development of an accurate understanding of the rela- tion between components of the system and provide information of a nature that is necessary for solving different types of problems. This appears to be particularly critical in extremely complex systems, such as the nuclear reac- tor (Woods, 1989~. When information from several sources must be integrated to perform a single operation, there are workload and performance benefits if the infor- mation requires the same type of processing and is displayed in an inte- grated format. Carswell and Wickens (1987) demonstrated that independent tasks are performed best with separate displays; integrated tasks are per- formed best with similar, integrated displays. Display integration was cre

WORKLOAD FACTORS 71 ated by depicting movement in different axes of control as variations in different dimensions of a single triangular object. Performance was also improved when displays with similar formats were grouped, by placing them in proximity to each other. In addition, when similar or related infor- mation is displayed on several instruments, performance can be improved by coding it in a similar manner (e.g., by color, shape, or numeric rather than pictorial format) (Boles and Wickens, 1987~. Sequentially presented information can be integrated by its temporal organization. Wickens et al. (1988) suggested that cockpit displays could be used more effectively if they were congruent with a pilot's mental model of the environment. Compatibility might be improved by the use of: (1) 3-dimen- sional perspective displays (rather than 2-dimensional planar displays); (2) color coding that corresponds with population stereotypes (e.g., red to indi- cate stop or slow down, green to go or speed up); and (3) consistent color coding of similar information across displays to aid integration. Workload will also increase if there is a lack of consistency across displays (Andre and Wickens, 1992~. Inconsistent formats should be avoided whenever possible. There are times, however, when changes in format are necessary and inevitable. For example, when an operator must cross-check between a fixed north-up map and a forward field of view out the cockpit window (see Chapter 7), or between a small-scale and large-scale map. In these cases, there are principles of visual momentum that effectively link the different displays together, showing how information depicted in one relates to the other (Andre et al., 1991; Aretz, 1991; Wickens, 1992b; Woods, 19841. Perceptual workload is related to the operator's goal. For example, looking at a specific display to read its value and glancing at the display during a routine scan to detect anomalies require different types of percep- tual processes. Scanning known locations for information that is presented in a consistent format (e.g., checking an instrument) is less demanding than searching for any target that might appear at any location (e.g., scanning a radar display for potential threats) (Liu and Wickens, 1990~. Thus, the source of perceptual difficulties may be different: determining the exact value of a particular instrument may depend on legibility or character size; detecting a change in the location of a pointer may depend on its rate or magnitude of change. Information From the Visual Scene In the real world, operators often rely on visual cues in the external scene (i.e., optical flow, structural transformations, terrain features) to esti mate their orientation, speed, altitude, heading, and location. They may have to do so because their instruments are not sufficiently accurate to perform a particular task (e.g., helicopter instruments are unreliable at very

72 WORKLOAD TRANSITION low altitudes or slow speeds) or because the information is not available (e.g., there is no compass in a tank). Perceiving spatial relationships and states is difficult if available visual cues are insufficient. For example, changes in optical flow may be too slow to provide useful guidance information at very slow speeds (Bennett et al., 1990~. In addition, variations in optical variables may be misinterpreted, even though they are perceptible. For example, Perrone (1984) found that pilots do not extract the appropriate cues from an external scene when estimating approach angle. Instead, they rely on simpler but less reliable cues. Kaiser et al. (1990) found that accurate perception of surface slope depends on the orientation of the observer's flight path relative to the ter- rain. Estimates are more accurate when the observer moves parallel to the slope than toward it. Compensating for such perceptual errors may result in a significant increase in workload during a critical phase of flight. At night, the crews of tanks and helicopters use visual aids (such as night vision goggles and thermal imaging systems) when monitoring the environment. Although these devices do provide information that would otherwise be unavailable, visual acuity, depth cues, and field of view are limited. Thus, perceiving objects and distances is more difficult, and pe- ripheral motion cues are unavailable or difficult to obtain. In addition, object perception with thermal imaging systems is considerably more diffi- cult than with direct vision. The visual display represents thermal differ- ences rather than reflected light. Thus, objects may look very different than they do with the naked eye (Foyle et al., 19901; surface texture and shading are limited or inconsistent with the viewer's expectations; shadows, convex, and concave areas may be either bright or dark depending on the polarity selected; and objects may disappear if their temperature is similar to that of their surround. Again, the perceptual demands thus imposed add signifi- cantly to operator workload. A different but related visual factor that may affect performance and workload is the use of spectacles or contact lenses. The National Research Council's Committee on Vision (1990; Flattau, 1991) has documented the advantages and disadvantages of their use in military aviation. These concerns should also be taken into consideration in this environment. In either monitoring or search tasks, attention is directed toward differ- ent positions in space, particular information sources, or both. In monitor- ing, the operator routinely scans one or more sources to detect the presence of a previously specified signal or a change in the value of a particular parameter. This task involves both temporal uncertainty (i.e., signals occur infrequently and at unpredictable times) and spatial uncertainty (i.e., a sig- nal may occur in any one of many spatial locations or information sources). In addition, monitoring continuously varying signals is more difficult than detecting the occurrence of a discrete event. In the former, signals may not

WORKLOAD FACTORS 73 be uniquely specified and the operator must infer a change in system state by integrating information from multiple sources (Wiener, 1984~. The time required to scan an array of instruments, and the demands thus imposed, are influenced by the number of different fixations required. Moving the eyes from one fixation to another takes from 70 to 700 milliseconds (Card et al., 1986~. A detailed discussion of the workload and performance effects of monitoring is presented in Chapter 6. Information From Auditory Displays Auditory information may be presented as aural displays (warning tones, beeps, synthetic voice messages) or voice communications among the crew or with people located elsewhere. The increasing use of aural warnings (a proliferation of beeps, buzzes, tones, etc.) in aviation demands increasing perceptual accuracy to distinguish among the possibilities. Environmental noise may create significant perceptual problems, particularly in tanks or helicopters. Furthermore, transmission quality may make it difficult to perceive a radio or intercom. Finally, many communications channels are filled with messages directed toward others. Thus, crews must monitor messages intended for other crews (in order to maintain situational aware- ness) while identifying and paying particular attention to messages intended for them. An aural message lasts for a finite amount of time, and elements of the message are presented sequentially. Thus, auditory messages have a certain attention grabbing quality because they must be perceived as they occur. If operators delay, they may miss important information. Furthermore, long messages may impose additional memory demands. Perceiving individual components is only the first step of perceiving the meaning of the whole message. (Speech communication is discussed further in Chapter 10.) Regional accents or computer-generated speech may increase the workload of speech perception. For example, the use of synthesized, rather than natural, speech for warning messages is becoming more common. Although the artificial sound of the voice commands the operator's attention, specific perceptual problems are created by speech quality (e.g., the pilots of trans- port aircraft reported significant difficulties interpreting synthesized voice messages used to convey critical warnings). Thus, auditory workload is increasing as quickly as visual workload in many complex systems. For example, civil and military pilots must monitor an increasing number of communication frequencies to maintain coordina- tion, as well as a variety of aural warnings. Soon auditory workload is likely to become as significant a workload problem as visual workload. Because the visual and auditory demands imposed on the operators of many complex systems are so high, and because incoming information may

74 WORKLOAD TRANSITION be degraded or difficult to interpret, problems associated with the initial perception of needed information impose additional demands on later cog- nitive processes required to identify, interpret, and integrate the informa- tion. Despite the likelihood that a relationship exists between perceptual difficulty and workload, relatively little workload research has focused on purely perceptual factors. Instead, the focus has been on the processes required to interpret, transform, and remember information and formulate a response. Information Processing Variables As the cognitive processing required to interpret incoming information and select a response increases from simple rules of thumb to complex algorithms, inference, or deduction, response time generally increases; er- rors are more frequent; and experienced workload is higher. Level of Processing Different factors affect the processing required at one or more stages of human information processing. For example, signal quality affects the ini- tial processing stage, and uncertainty affects the response selection stage (Gopher and Donchin, 1986J. Craik and Lockhart (1972) suggested that preliminary analyses focus on the physical features of a stimulus, while later stages are more concerned with matching the input against information stored in memory, extracting meaning, and enrichment or elaboration. Fa- miliar, meaningful input, which is compatible with existing cognitive repre- sentations, is processed at a deeper level more quickly than is unfamiliar, or less meaningful, input. Within this framework, workload might be linked to the level and depth of analysis required to interpret and retain information. Rasmussen (1982) identified three categories of tasks, differentiated by the level of processing required for their performance. Levels of processing were associated with degrees of uncertainty. The greater the uncertainty associated with selecting an appropriate response for a particular situation, the higher the cognitive demands placed on an operator. For skill-based tasks, which are exemplified by highly practiced perceptual-motor skills, there is a clear and unambiguous relationship between system states and required responses, and no uncertainty about the mapping from stimulus to action. Rule-based tasks are characterized by a set of appropriate actions governed by explicit procedures. Once an operator has correctly recognized the situation, the choice of actions is deterministic, following a set of if- then rules. Knowledge-based tasks are characterized by uncertainty, re- quirements to develop novel solutions, and delayed or limited feedback. These activities require operators to perform complex interpretation and

WORKLOAD FACTORS 75 decision making. Although tasks may be readily assigned to Rasmussen's three categories, recent evidence suggests that their ordering of skill-rule- knowledge does not necessarily correspond to their level of workload. For example, Moray et al. (1988) found that skill-based tasks did impose sig- nificant workload on people performing a complex video-game-like labora- tory task. In fact, subjective workload estimates were often dominated by the difficulty of the skill-based component of the task. Since there is less uncertainty associated with skill- and rule-based behaviors, these tasks are often candidates for automation. Knowledge- based tasks are left for the human operator, as they are difficult or impos- sible to automate. While it is in this domain that humans continue to outperform machines, these functions are more difficult to learn, easier to forget (as they are often quite abstract), and impose high workload (Hart and Sheridan, 19849. In support of this concept, Tanaka et al. (1983) dem- onstrated that automation did in fact reduce the workload of skill-based tasks, but not rule- and knowledge-based tasks, in a laboratory flight simu- lation. However, unexpected and abnormal events increased the workload of rule- and knowledge-based tasks. Closely related to the skill-rule-knowledge continuum is the dichotomy between automatic and controlled processing. If there is a unique and consistent mapping between signals and actions, rapid and automatic pro- cessing develops with experience (Schneider and Shiffrin, 1977~. Process- ing is fast, inflexible, difficult to suppress (once learned), and not limited by short-term memory capacity or attention. Workload is generally low. Similar improvements may be found even when stimulus-response relation- ships vary, provided the rules relating types of stimuli to appropriate re- sponses are consistent (Kramer et al., 19881. When the relationship be- tween signals and responses is not consistent, automatic behavior cannot develop. Processing remains serial and capacity-limited in the mode known as control processing. The higher-order mental processes required to select a response increase response time and impose higher workload. Thus, a consistent relationship between the location, appearance, and meaning of a particular display element, or class of features, will improve performance and reduce workload. As an example, automatic processing will be more likely to develop with a fixed display that has a fixed mode of information presentation than a variable mode or multifunction display. Processing Resources More recently, performance limitations have also been linked to the availability of resources (i.e., hypothetical entities that represent the capac- ity or mental energy available to perform a task). Early models (see, for example, Kahneman, 1973), assumed that resources were limited in quantity

76 WORKLOAD TRANSITION but relatively nonspecific. Performance suffers if any task or combination of tasks requires more capacity than is available. Evidence for an undiffer- entiated pool of resources came from the observation that apparently unre- lated activities interfere with each other's performance. Kahneman further suggested that the arousal induced by some task requirements might in fact increase available capacity. Although common experience supports the no- tion that people can perform incredible feats when highly motivated, the concept of flexible capacity proved to be difficult to test. The common observation that tasks interfere with each other to differ- ent degrees suggests the existence of independent resources for perceiving, encoding, processing, and responding to different types of information. The seminal papers written by Norman and Bobrow (1975) and Navon and Go- pher (1979) resulted in hundreds of experiments designed to identify and describe such resources. Most of these studies used a dual-task paradigm in which pairs of tasks were presented that imposed various demands on dif- ferent resources. Wickens's (1980) multiple resource model proposed the following dimensions: input modality (visual/auditory), processing stage (per- ceptual-cen~al/response), processing code (spatial/verbal), and response modality (manual/verbal). The relationship between performance in single- and dual- task conditions was interpreted as evidence of competition for the same resource or independent use of different resources. The general finding has been that when the resource requirements of two tasks are the same (e.g., visual perception, spatial processing, manual response) and exceed an operator's capacities, then response latency, accu- racy, or both on one or both tasks will suffer. But if, for example, one task is presented auditorially (rather than visually) or requires verbal (rather than spatial) processing or a vocal (rather than manual) response, it is more likely that both tasks can be performed as well together as separately. For example, Vidulich and Tsang (1985) found that performance was better and workload lower when a visual-spatial-manual tracking task was paired with a spatial transformation task presented auditorially and performed verbally than with a visual-manual version. Similar results were found in a more complex environment. Vidulich and Bortolussi (1988) conducted two simu- lations of advanced Army scout/attack helicopter missions to evaluate alter- native methods of entering discrete commands. Although commands took longer to enter using a voice recognition system than with manual switches (and subjective workload was higher), single- to dual-task performance dec- rements were less for voice commands than manual. Furthermore, simulta- neous helicopter control performance was better with voice commands. Manual discrete tasks competed with manual flight control for limited resources, whereas vocal discrete tasks did not. Wickens's (1980, 1991) multiple resource model, based on extensive experimental data, provides a useful tool for the designers of complex sys

WORKLOAD FACTORS 77 tems. This model predicts the time-sharing efficiency likely for different task combinations based on their resource-demand profile (see, for example, Wickens, 1989~. It provides a framework for evaluating the effects of alternative design concepts on operator performance and improving the hu- man-machine interface. If an operator's visual capabilities are exceeded by ongoing activities, then alternative display modes might be considered. If control activities exceed an operator's manual capabilities, vocal controls might be substituted for some tasks, and so on. Although the research performed by Wickens and his colleagues clearly demonstrated the performance problems associated with specific combina- tions of task parameters, the relationship between resource concepts and workload, at least subjective estimates of workload, was less clear (see, for example,Vidulich end Wickens, 1986; Yeh end Wickens, 1988~. Although humans appear to be aware of the total demands imposed on them (created by variations in task frequency, difficulty, and number), their estimates of workload are relatively insensitive to resource competition. If this is true, then it presents the operator of a complex system with a significant prob- lem. If humans are not aware of the potentially detrimental effects of specific task combinations on performance, then they might not adopt the most efficient strategies for performing particular combinations of tasks that momentarily exceed their capacities. Memory Requirements Most models of the human mind distinguish among different types of memory: (1) modality-specific sensory registers (in which auditory and visual stimuli are represented literally for no more than a few seconds); (2) some type of resource-demanding short-term or working memory (in which a limited number of items are held for seconds or minutes); and (3) long- term memory (in which vast quantities of information are held for long periods of time). Although attention is not required for information to enter sensory registers, the information will be lost by the process of displace- ment unless it is transferred into working memory for further processing (Neisser, 1976~. Short-term memory is used to temporarily hold and ma- nipulate new information or that retrieved from long-term memory. Only some of the information once held in short-term memory is transferred to long-term memory for more permanent retention. It is unlikely that the products of preliminary or intermediate analyses are retained. Craik and Lockhart (1972) suggested that the persistence of information in memory, and the way it is encoded, depend on the level or depth of processing it received initially. Because there are severe limitations in the number of separate items that can be retained in working memory at any point in time, errors are

7g WORKLOAD TRANSITION more likely for tasks that exceed this capacity (Card, 19&9~. Miller (1956) placed this limit at seven items, plus or minus two. However, he also suggested that chunking related information could increase this apparent limitation. Chunks were described as composite units created by grouping, organizing, or recoding a group of otherwise separate elements based on rules stored in long-term memory. Although a useful concept, the possibil- ity of chunking eliminates the feasibility of assessing the processing de- mands of tasks based solely on their formal properties or surface structure. Information decays rapidly from short-term memory, and not all infor- mation is retained equally well. For example, the last items in a list are generally recalled most accurately (i.e., recency effect), followed by the first items in a list (i.e., primacy effect) (Postman and Phillips, 1965~. A1- though information can be retained in working memory almost indefinitely by active rehearsal (Baddeley, 1986), some or all of it may be lost if new information or higher-priority tasks intervene (i.e., it is difficult to remem- ber a telephone number while also engaging in a conversation). However, there may be different types of working memory capable of retaining and processing different types of information. For example, Baddeley and Hitch (1974) proposed two relatively independent processors: one that stores and manipulates verbal information, and another that stores and manipulates visual-spatial images. As the demands imposed on short-term memory are increased, tasks are perceived as being more difficult. Thus, tasks that require short-term memory often impose greater workload than do tasks that require retrieval of infor- mation from long-term memory. For example, Berg and Sheridan (1984) found that pilot workload was significantly higher and flying performance was significantly worse for flights that imposed short-term memory de- mands than for those that required long-term memory. Since memory is fallible, particularly when interruptions occur, written procedures and checklists are used in many operational environments to reduce operator workload and decrease the probability of errors. However, since retrieving the appropriate page from a manual may take an unaccept- able amount of time in an emergency situation, critical items (e.g., the first few steps in an emergency checklist) are committed to memory. This en- sures a rapid and appropriate response when stress or time pressure inter- fere with operators' analytical and decision-making abilities. Alternatively, people may deliberately bring information that they anticipate needing into working memory, where it will be readily accessible. They recognize that it takes time to access information from long-term memory and that they may forget important information under time stress. A series of actions that often occur together may become integrated into a high-level action sequence and represented in memory as a single event. However, if information is remembered as an element of an ordered

WORKLOAD FACTORS 79 list, or if a series of actions are represented as an integrated unit, it is more difficult to retrieve an individual element of this series from memory. Just because people possess relevant knowledge in long-term memory does not guarantee that this knowledge will be activated when needed (Woods, 1989~. Display-Control Compatibility The way in which information displays are mapped onto the appropriate responses influences both workload and performance. In general, responses are faster and more accurate when there is a spatial correspondence between the position and direction of movement of stimuli and control responses, correspondence with population stereotypes, compatibility between stimu- lus and response modalities, and agreement of control dynamics with task requirements (Jagacinski, 1989~. As one example, the traditional relation- ship between helicopter controls and displays places the instrument for alti- tude (which is controlled with the left hand) on the right and for speed (which is controlled with the right hand) on the left. Craig et al. (1983) found that a more compatible arrangement resulted in faster movement times than did the traditional arrangement (e.g., responses were 1 to 2 seconds faster with a compatible arrangement). The compatibility between display modality, response modality, and type of processing required to perform a task also influences performance and workload (Vidulich and Wickens, 1986; Wickens et al., 19831. For both discrete and continuous control tasks, single-task performance and time- sharing efficiency are better when display modality is compatible with: (1) the way information is cognitively represented and processed (e.g., visual displays are more compatible with spatial processing; auditory displays are more compatible with verbal processing) and (2) response modality (e.g., auditory displays are more compatible with vocal responses; visual displays are more compatible with manual responses). Since incompatible arrange- ments impose additional information processing demands, it is reasonable to assume they should increase workload, as well as slow performance (Andre and Wickens, 1992~. If information must be transformed from one format or orientation to another to complete a task, workload and response time are increased. For example, King et al. (1989) found that response time and subjective workload increased when mental rotation was required to bring two items into align- ment for comparison. Aretz (1990) found, as others have, that the time required to mentally rotate a map to bring it into alignment with the exter- nal scene increased from 3.7 to 4.2 seconds as the difference in orientation increased from O to 180 degrees. If information must be transformed from one form to another to per- form sequential tasks, response time is longer and subjective workload is

80 WORKLOAD TRANSITION higher than if the two tasks share common input modalities and processing codes. For example, Hart et al. (1986) found that visual-spatial target ac- quisitions took longer and imposed higher workload when the target to be acquired was identified by auditory rather than visual displays or by ver- bally coded rather than spatially coded commands. These and other results suggest that unrelated but concurrent tasks ben- efit from displays that require different sensory modalities, processing codes, and response modalities. However, performance on functionally related, sequential tasks benefits from compatible display modalities, central pro- cessing codes, and response modalities. The former reduces competition for limited resources (thereby reducing workload and improving perfor- mance), whereas the latter promotes subjective integration and reduces the need for transformation (also reducing workload and improving performance). Output Variables Although the intrinsic difficulty of a task determines the demands im- posed on an operator, the interface that is provided to accomplish the task has a significant influence on the workload actually experienced. Poorly designed controls, high-order system dynamics, inadequate displays, and incompatible controls and displays may make it difficult for an operator to accomplish even relatively easy tasks. Control Design Many real-world tasks require discrete, transient, periodic, or relatively continuous control inputs in response to direct or indirect feedback from the environment. Pilots move the control yoke to change heading or altitude, drivers steer their vehicles down a road, and gunners track and acquire a target. The workload of these activities depends on the independent and joint effects of: (1) the frequency, duration, and force of control inputs required to change (e.g., complete a turn, move the reticle over the target) or maintain the state of the system (e.g., minimizing the effects of distur- bances created by uneven road surface, wind gusts), (2) the way control inputs map onto a change in system state, (3) the way changes in system state are displayed, (4) the physical characteristics and dynamics of the control devices), (5) the number of axes controlled, (6) the availability of predictability and preview, and (7) the required precision. Although control tasks may impose relatively constant physical demands (i.e., operators' hands and feet are busy), cognitive demands may be low (i.e., perceptual-motor tasks may require little thought for an experienced operator, such as a back- hoe operator). Manual control workload is influenced by the physical and dynamic

WORKLOAD FACTORS 81 characteristics of the control device itself and of the system controlled. There may be one (or many) control devices used to operate a system. A single control input may impact several systems, or coordinated use of several controls may be required to make a single change in system state. Control devices may be single- or multiaxis. The more complex the control system, the greater the workload of the operator. For example, the more axes that must be controlled simultaneously, the higher the workload of pilots in simulated flight (Kantowitz et al., 1984; King et al., 1989~. For most vehicle control tasks, workload and performance are influenced by interrelationships among axes of control. For example, if a helicopter pilot pulls back on the cyclic to gain altitude, he must simultaneously add power (with the collective) to maintain a constant speed. If he banks to the right to execute a turn, he must increase pressure on the pedals to counteract the additional torque. These complex interrelationships among control devices, control surfaces, and vehicle state impose additional cognitive load on op- erators. Control Gain and Display Gain Control gain refers to the displacement or force required to achieve a given change in system state. In high-gain systems, small control move- ments produce large outputs. In low-gain systems, large control movements are required to achieve the same effect. Display gain refers to the way in which changes in system-state are presented to the operator. A high-gain display reflects even minor variations in system state; while, a low-gain display reflects only those changes that exceed some threshold. Although humans are able to compensate for variations in control and display gain over a wide range (Wickens, 1986), additional workload is imposed when control or display gain exceeds an optimal range, is poorly matched, or is inappropriate for a specific task. If gain is too high, the effects of even small corrections are magnified and the operator must make very fine ad- justments to avoid introducing errors. Oscillations and overcorrections are often the result. If gain is too low, frequent or large-amplitude movements may be required, inducing physical fatigue. While high-gain systems re- duce the workload of the initial phase of target acquisitions (the higher the gain, the smaller the control displacement required), they impose additional workload during the final phase (the higher the gain, the finer the correc- tions required to remain over the target area). Control Lag and Display Lag A lag in the system's response to a control input can increase operator workload, although a pure time delay is not particularly disruptive when

82 WORKLOAD TRANSITION inputs are predictable (Wickens, 19861. Generally, as the delay between a control input and a change in the display of system state increases, tracking error and workload increase. Levison et al. (1979) found that performance was disrupted by delays as small as 80 milliseconds between simulator motion and visual displays. Preview can improve performance and reduce workload; that is, when the operator can see ahead and anticipate future control requirements (e.g., view turns in the road ahead, predict periodicities in disturbances) operators can shift from a reactive to a proactive mode of behavior. Time lags coupled with high gains produce dynamics that are likely to be unstable and oscillatory. Order of Control Control dynamics generally fall into one of three categories (or are a combination of more than one): (1) Zero-order or position controls are the easiest to conceptualize and use. A deflection of the control stick causes the controlled element to move in the direction the stick was deflected. If the stick is held steady in one position, the controlled element remains stationary. (2) First-order or velocity controls command rate. If stick de- flection is constant, velocity and direction are constant; the greater the stick deflection, the faster the controlled element moves in that direction. (3J Second-order or acceleration controls command rate of change; the greater the stick deflection, the faster the controlled element accelerates in that direction. If stick deflection is held constant, the controlled element will move in the commanded direction at an increasing velocity. Although tracking performance and workload are not generally differ- ent between zero- and first-order controls, they differ significantly between first- and second-order controls (Vidulich and Wickens, 1984; Wickens, 19861. The difficulty of higher-order control is attributed to the amount of lead, or prediction, an operator must generate to achieve the same level of performance, the information processing required, and the complexity and number of inputs necessary to achieve the same result. For target acquisi- tion, as with continuous tracking, the higher the order of control, the greater the demands placed on the operator. Put very simplistically, position, ve- locity, and acceleration controls require a minimum of one, two, or three control inputs to bring and hold a reticle over a target. Computer Aiding and Automation Computer aiding and automation are introduced to reduce operator workload, increase safety, improve performance, and extend mission capability. In automobiles, automatic transmission and cruise control reduce the need for gear changes and continuous operation of the gas pedal. In transport air

WORKLOAD FACTORS 83 craft, autopilots achieve and maintain the flight path based on information entered by the pilot before takeoff or modified inflight by single, discrete commands. In military jets, terrain following/terrain avoidance algorithms process mission, terrain, and threat data to project and follow an optimal flight path. In many environments, even system monitoring is accomplished automatically, providing the operator with visual and/or auditory alarms when a failure occurs. Expert systems are being developed for use in high- performance aircraft that will infer the pilots' intentions and evaluate the degree to which his commands support, or conflict with, his current goals (Rouse et al., 1990~. Given the increasing complexity of many advanced systems, operators are no longer able to control the system without assistance. Thus, various methods of alleviating their workload have been introduced. For example, direct displays of system parameters (which may be noisy, high-order, and unintegrated) might be replaced by displays that filter noisy inputs, appear to be of a lower order, and integrate the outputs of related subsystems. Predictor displays can depict estimates of the future state or position of the system based on its current state and assumptions about the operator's fu- ture control activity (Wickens, 1986) and present this information to the operator in graphic form. Because predictor displays perform some integra- tion and projection for operators, thus allowing them to be proactive rather than reactive, they often improve the accuracy and smoothness of control activities and reduce operator workload. Stability augmentation systems reduce the frequency and order of con- trol inputs required of an operator. For example, a number of the stability- control augmentation systems evaluated in a simulated advanced Army heli- copter significantly reduced pilot workload (Haworth et al., 1987), but the benefits varied across mission segments. Fully automatic systems allow an operator to enter a desired outcome directly (e.g., a new heading) while underlying subsystems manipulate the control surfaces to achieve the de- sired state. In fact, as the level of control augmentation increases beyond some point, the role of the operator shifts from that of manual controller to system monitor. Vienneau and Gozzo (1987) evaluated the effects of automating eight different subsystems in simulated military missions conducted in a single- pilot advanced scout/attack helicopter. They found, as did Haworth et al. (1987), that automation aided pilots in managing workload peaks and re- duced overall workload. For example, computer-aided systems for commu- nications (with voice-activated frequency selection) and navigation (with a digital map display) reduced flight path deviations by an average of 67 percent and time en route by 70 percent. Again, differential benefits were found across mission segments and greater improvements were found for high-workload activities. Vienneau and Gozzo suggested that workload

84 WORKLOAD TRANSITION reductions might be achieved by either the addition of automated subsystems or by increasing the efficiency of the human's interface with the system. Automation does not always reduce operator workload. Often, it sim- ply replaces one form of workload (the physical demands of manual con- trol) with another (the perceptual-cognitive demands of monitoring the sys- tem) (Hart and Sheridan, 1984~. Designers tend to automate tasks that are easy to automate. As discussed earlier, these are usually the skill-based tasks that are also easy for a human to perform. This often leaves the operator with a collection of unrelated activities to perform that are too ambiguous or unpredictable to automate and that impose equally high de- mands on the operator. Finally, automation reduces an operator's direct involvement with and knowledge of the system and may make it more difficult to detect errors or assume control if the automated system fails. For example, Wickens and Kessel (1979) found that people are better able to detect a sudden change in system state when manually controlling a system than when supervising the performance of an automated system. Thus, opportunities for system error may, in fact, increase. Long delays between an operator input and its ultimate effect decrease the probability of detecting and rectifying a human error (Kantowitz and Casper, 1988) or a system error. Although airline pilots have found automation to be useful, they do not necessarily agree that it has reduced their overall workload (Wiener, 1989; Wiener and Curry, 1980~. In many cases, automation is introduced to allow a reduction in crew size or an increase in mission capability. This creates the potential for increasing the workload of remaining crew members if the level of automa- tion is not sufficient. Even with a fully functioning system, it is difficult to replace a human operator completely with an automatic system. For ex- ample, Haworth et al. (1987) found it was impossible to achieve the same level of workload typical of two-pilot crews for the single pilots of a highly automated simulation of an advanced Army helicopter. In addition, workload and performance predictions and system evaluations (on which crew complement decisions are based) assume the appropriate use of available automation. If the systems are too numerous or complex to be used effectively, or if they fail, the burden of accomplishing the mission rests on the operators, who may experience far higher workload than anticipated. Finally, the contro- versy continues about how to allocate tasks between humans and machines and whether the operator should retain control or allow an automated sys- tem to override human decisions. SUMMARY In some cases, it is apparent that human limitations reflect the conse- quences of poorly designed controls, displays, and automatic subsystems.

WORKLOAD FACTORS 85 In others, task demands simply exceed the operator's capabilities either momentarily or for extended periods. Despite their limitations, humans are remarkably flexible, adaptable, and capable. They can improvise, compen- sate for inadequate information and system or human failures, adjust to novel situations, exhibit graceful (rather than catastrophic) degradation, plan ahead, predict the outcome of familiar and unfamiliar events, and learn from experience. However, the consequences of extreme demands and re- quirements to act creatively and adaptively impose significant workload on the human operators of complex systems. Regardless of the specific sourcegs) of workload at any point in time, adequate training and preparation, adopting strategies and tactics most ap- propriate for the situation, effective leadership, and smooth crew coordina- tion can counteract some of the detrimental effects of imposed task de- mands, transitioning from one mode of behavior to another, environmental stressors, and fatigue. The more actions that have become automatic through training and experience and the more predictable events seem through plan- ning and rehearsal or availability of preview information, the more likely it will be that crews will respond appropriately and sustain adequate perfor- mance for as long as necessary. There are several significant factors that characterize the transition from low to high workload mission phases. One, of course, is the absolute level of workload. There is no question that workload is higher during an en- gagement than before. However, the fact that the sources of workload are significantly different between the two phases may present a greater prob- lem. The nature of the task demands shifts from preparation to action; from static to dynamic; from long lead times to short; from passive monitoring and maintenance to active information seeking, control, and operating; from direct interaction to radio communications; from predictive to compensa- tory behaviors; from planning to reaction; from organized procedures and schedules to spontaneous actions. Crews may adopt different strategies for coping with a sudden increase in workload. They may choose to process fewer events (e.g., selectively attend to fewer tasks, defer activities, monitor fewer channels, ignore infor- mation about expected events, or consider fewer alternative hypotheses). Alternatively, they may choose to process events less completely (e.g., sample information sources less often, seek less corroborating evidence, be satis- fied with partial feedback, narrow the field of attention, pursue fewer pos- sible explanations, or limit their anticipation of what might happen next) (Woods, 19891. Although either of these strategies might be effective in reducing workload in the near term, they also might shift such workload to a later time. Furthermore, incomplete attention to the current situation and tasks is more likely to result in errors, which will impose additional workload to rectify.

86 WORKLOAD TRANSITION A shift in the nature of task demands, accompanied by an increase in time pressure and uncertainty, requires a significant change in the team's behaviors and strategies, which in itself imposes additional workload. In addition, boredom, impatience, or apprehension (which may characterize the premission phase) are replaced by stress and fear following a transition. While the latter factors increase the level of arousal and focus the attention of the crew (a potential benefit for performance) they may impair the crew's ability to perform accurately and effectively, thereby increasing the effort it must exert to maintain acceptable performance. These stress effects are addressed in the next chapter. REFERENCES Acton, W.H., M.S. Crabtree, J.C. Simons, F.E. Gomer, and J.S. Eckel 1983 Quantification of crew workload imposed by communications-related tasks in commercial transport aircraft. Pp. 239-243 in Proceedings of the Human Factors Society 27th Annual Meeting. Santa Monica, California: Human Factors Society. Andre, A.D., and C.D. Wickens 1992 Compatibility and consistency in display-control systems. Human Factors 34(6). Andre, A.D., C.D. Wickens, L. Moorman, and M.M. Boeschelli 1991 Display formatting techniques for improving situation awareness in the aircraft cockpit. International Journal of Aviation Psychology 1(3):205-218. Aretz, A.J. 1990 Map display design. Pp. 89-93 in Proceedings of the Human Factors Society 34th Annual Meeting. Santa Monica, California: Human Factors Society. Baddeley, A.D. 1986 Working Memory. Oxford: Clarendon Press. Baddeley, A.D., and G.J. Hitch 1974 Working memory. In G.H. Bower, ea., The Psychology of Learning and Motiva- tion: Advances in Research and Theory, Volume 8. New York: Academic Press. Bamba, Z., D. Bushnell, S. Chen, A. Chiu, C. Nuekom, S. Nishimura, M. Prevost, R. Shankar, L. Staveland, and G. Smith 1991 Army-NASA Aircrew Aircraft Integration Program (A3I): Man-Machine Integra- tion, Design, and Analysis Systems (MIDAS). Report No. TN-91-8216-000. Palo Alto, California: Sterling Federal Systems, Inc. Bennett, C.T., M. Schwirzke, and J.S. Tittle 1990 Perceptual and Performance Consequences of Flight in Virtual Worlds. Poster presented at the Workshop on Human-Machine Interfaces for Teleoperators and Virtual Environments. Santa Barbara, California. Berg, S.L., and T.B. Sheridan 1984 Measuring workload differences between short-term memory and long-term memory scenarios in a simulation flight environment. Pp. 397-416 in Proceedings of the 20th Annual Conference on Manual Control. Report No. NASA CP-2341. Wash- ington, DC: National Aeronautics and Space Administration. Bittner, A., J.C. Byers, S.G. Hill, A.L. Zaklad, and R.E. Christ 1988 Generic Workload Ratings of a Mobile Air Defense System (LOS-F-H). Willow Grove, Pennsylvania: Analytics, Inc. Boles, D.B., and C.D. Wickens 1987 Display formatting in information integration and nonintegration tasks. Human Factors 29(4):395-406.

WORKLOAD FACTORS 87 Bortolussi, M.R., S.G. Hart, and R.J. Shively 1987 Measuring moment-to-moment pilot workload using synchronous presentations of secondary tasks in a motion-base simulator. Pp. 625-657 in Proceedings of the Fourth Symposium on Aviation Psychology. Columbus, Ohio: Ohio State Univer- sity. Broadbent, D.E. 1990 Fellows profile: Donald Broadbent. Pp. 7 in The Human Factors Society Bulletin. Santa Monica, California: Human Factors Society. Card, S.K. 1989 Models of working memory. Pp. 203-214 in J. Elkind, S. Card, J. Hochberg, and B. Huey, eds., Human Performance Models for Computer-Aided Engineering. Washington, DC: National Academy Press. Card, S.K., T.P. Moran, and A. Newell 1986 The model human processor. Pp. 45.1-45.35 in K. Boff, L. Kauffman, and J.P. Thomas, eds., Handbook of Perception and Human Performance. New York: Wiley and Sons. Carswell, M., and C.D. Wickens 1987 Information integration and the object display: An interaction of task demands and display superiority. Ergonomics 30:511-528. Craig, K.M., E.J. Hartzell, and S.L. Dunbar 1983 Helicopter pilot responses as a function of compatibility of the control-display configuration. Pp. 100-115 in Proceedings of the 19th Annual Conference on Manual Control. Washington, DC: National Aeronautics and Space Administra- tion. Craik, F.I., and R.S. Lockhart 1972 Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior 11:671-684. Cumming, R.W., and P.G. Croft 1973 Human information processing under varying task demand. Ergonomics 16:581- 586. Flattau, P.E., ed. 1991 Consideration in Contact Lens Use Under Adverse Conditions: Proceedings of a Symposium. Committee on Vision, National Research Council. Washington, DC: National Academy Press. Foyle, D.C., M.S. Brickner, L. Staveland, and B.D. Sanford 1990 Human object recognition as a function of display parameters using TV and infra- red imagery. Proceedings of the Society for Information Displays 1990 Interna- tional Symposium. Gopher, D., M. Brickner, and D. Navon 1982 Different difficulty manipulations interact differently with task emphasis: Evi- dence for multiple resources. Journal of Experimental Psychology: Human Per- ception and Performance 8(1):146-157. Gopher, D., and E. Donchin 1986 Workload: An examination of the concept. Pp. 41.1-41.49 in K. Boff, L. Kauffman, and J.P. Thomas, eds., Handbook of Perception and Human Performance. New York: Wiley and Sons. Gopher, D., M. Weil, T. Bareket, and S. Caspi 1988 Fidelity of task structure as a guiding principle in the development of skill trainers based upon complex computer games. Pp. 1266-1275 in Proceedings of the Hu- man Factors Society 32nd Annual Meeting. Santa Monica, California: Human Factors Society.

88 WORKLOAD TRANSITION Gopher, D., M. Well, and D. Siegel 1989 Practice under changing priorities: An approach to training complex skill. Acta Psychologica 71(1-3):147-177. Hahler, B., S. Dahl, K.R. Laughery, Jr., J. Lockett, III, and B. Thein 1991 CREWCUT: A tool for modeling the effects of high workload on human perfor- mance. Pp. 1210-1214 in Proceedings of the Human Factors Society 35th Annual Meeting. Santa Monica, California: Human Factors Society. Hancock, P.A., and J.S. Warm 1989 A dynamic model of stress and sustained attention. Human Factors 31:519-537. Hanson, R.H., D.G. Payne, R.J. Shively, and B.H. Kantowitz 1981 Process control simulation research in monitoring analog and digital displays. Pp. 154-158 in Proceedings of the Human Factors Society 25th Annual Meeting. Santa Monica, California: Human Factors Society. Harris, R.L., Sr., J.R. Tole, A.R. Ephrath, and A.T. Stephens 1982 How a new instrument affects pilots' mental workload. Pp. 1010-1013 in Pro ceedings of the Human Factors Society 26th Annual Meeting. Santa Monica, California: Human Factors Society. Hart, S.G. 1986 Theory and measurement of human workload. Pp. 396-456 in J. Zeidner, ea., Human Productivity Enhancement: Training and Human Factors in System De sign, Volume I. New York: Praeger. 1989 Crew workload management strategies: A critical factor in system performance. Pp. 22-27 in Proceedings of the Fifth International Symposium on Aviation Psy chology. Columbus, Ohio: Ohio State University. Hart, S.G., V. Battiste, and P.T. Lester 1984a POPCORN: A supervisory control simulation for workload and performance re- search. Pp. 431-454 in Proceedings of the 20th Annual Conference on Manual Control, Volume 1. Report No. NASA-C-2341. Washington, DC: National Aero- nautics and Space Administration. Hart, S.G., and M.R. Bortolussi 1984 Pilot errors as a source of workload. Human Factors 26(5):545-556. Hart, S.G., and J.R. Hauser 1987 Inflight application of three pilot workload measurement techniques. Aviation, Space, and Environmental Medicine 58(5):402-410. Hart, S.G., J.J. Sellers, and G. Guthart 1984b The impact of response selection and response execution difficulty on the subjec- tive experience of workload. Pp. 732-736 in Proceedings of the Human Factors Society 28th Annual Meeting. Santa Monica, California: Human Factors Society. Hart, S.G., and T.B. Sheridan 1984 Pilot workload, performance, and aircraft control automation. Pp. 18.1-18.12 in Proceedings of the AGARD Symposium on Human Factors Considerations in High Performance Aircraft Conference Proceedings No. 371. Neuilly sur Seine, France: NATO Advisory Group for Aerospace Research and Development. Hart, S.G., R.J. Shively, M.A. Vidulich, and R.C. Miller 1986 The effects of stimulus modality and task integrality: Predicting dual-task perfor- mance and workload from single task levels. Pp. 5.1-5.18 in Proceedings of the 21st Annual Conference on Manual Control. Report No. NASA CP-2428. Wash- ington, DC: National Aeronautics and Space Administration. Hart, S.G., and C.D. Wickens 1990 Workload assessment and prediction. In H.R. Booher, ea., MANPRINT: An Emerging Technology. Advanced Concepts for Integrat~ng People, Machines, and Organiza- tions. New York: Van Nostrand and Reinhold.

WORKLOAD FACTORS 89 Hauser, J.R., M.E. Childress, and S.G. Hart 1983 Rating consistency and component salience in subjective workload estimation. Pp. 127-149 in Proceedings of the 18th Annual Conference on Manual Control. Report No. AFWAL-TR-8-3021. Wright-Patterson Air Force Base, Ohio: Air Force Flight Dynamics Laboratory. Haworth, L.A., A. Atencio, C.C. Bivens, R.J. Shively, and D. Delgado 1987 Advanced Helicopter Cockpit and Control Configurations for Helicopter Combat Missions. Report No. NASA TM-100017. Washington, DC: National Aeronau- tics and Space Administration. Hughes, D. 1989 Glass cockpit study reveals human factors problems. Aviation Week and Space Technology August 7:32-35. Hick, W.E. 1952 On the rate of gain of information. Quarterly Journal of Experimental Psychology 4:11-26. Hockey, G.R.J. 1986 Changes in operator efficiency as a function of environmental stress, fatigue, and circadian rhythms. Pp. 44.1-44.49 in K. Boff, L. Kauffman, and J.P. Thomas, eds., Handbook of Perception and Human Performance. New York: Wiley and Sons. Hurst, M.W., and R.M. Rose 1978 Objective job difficulty, behavioral responses, and sector characteristics in air route traffic control centers. Ergonomics 21(9):697-708. Hyman, R. 1953 Stimulus information as a determinant of reaction time. Journal of Experimental Psychology 45 :188- 196. Jagacinski, RJ. 1989 Target acquisition: Performance measures, process models, and design implica- tions. Pp. 135-150 in G. McMillan, ea., Applications of Models to System Design. New York: Plenum Press. Jagacinski, R.J., D.W. Repperger, S.L. Ward, and M.S. Moran 1980 A test of Fitts' Law with moving targets. Human Factors 22:225-233. Johnson, W.W., and S.G. Hart 1987 Step tracking shrinking targets. Pp. 248-252 in Proceedings of the Human Factors Society 31st Annual Meeting. Santa Monica, California: Human Factors Society. Kahneman, D. 1973 Attention and Effort. Englewood Cliffs, New Jersey: Prentice-Hall. Kaiser, M.K., J.S. Perrone, G.J. Anderson, J.L. Lappin, and D.R. Proffitt 1990 The Effect of Motion on Surface Slant Perception. Paper presented at the Annual Meeting of the Association for Research in Vision and Ophthalmology (ARVO). Sarasota, Florida. Kantowitz, B.H., and P.A. Casper 1988 Human workload in aviation. Pp. 157-187 in E. Wiener and D. Nagel, eds., Human Factors in Aviation. San Diego, California: Academic Press. Kantowitz, B.H., S.G. Hart, and M. Bortolussi 1983 Measuring pilot workload in a motion-based simulator: I. Asynchronous second- ary choice-reaction task. Pp. 359-372 in Proceedings of the Human Factors Soci- ety 27th Annual Meeting. Santa Monica, California: Human Factors Society. Kantowitz, B.H., S.G. Hart, M.R. Bortolussi, R.J. Shively, and S.C. Kantowitz 1984 Measuring pilot workload in a moving-base simulator: II. Building levels of load. Pp. 359-372 in Proceedings of the 20th Annual Conference on Manual Control.

9o Keele, S.W. WORKLOAD TRANSITION Report No. NASA CP-2341. Washington, DC: National Aeronautics and Space Administration. 1986 Motor control. Pp. 30.1-30.60 in K. Boff, L. Kauffman, and J.P. Thomas, eds., Handbook of Perception and Human Performance. New York: Wiley and Sons. King, T., J. Hamerman-Matsumoto, and S. Hart 1989 Dissociation revisited: Workload and performance in a simulated flight task. Pp. 796-801 in Proceedings of the Fifth International Symposium on Aviation Psychol- ogy. Columbus, Ohio: Ohio State University. Kramer, A., D. Strayer, and Y. Liu 1988 The role of consistency in automatic processing. Pp. 266-270 in The Proceedings of the Eleventh Biennial Psychology in the Department of Defense Symposium. Colorado Springs, Colorado: U.S. Air Force Academy. Levison, W.H., R.E. Lancroft, and A.M. Junker 1979 Effects of simulator delays on performance and learning in a roll-axis tracking task. Proceedings of the Fifteenth Annual Conference on Manual Control. Report No. AFFDL-TR-7-3 134. Wright-Patterson Air Force Base, Ohio: Air Force Flight Dynamics Laboratory. Liu, Y., and C.D. Wickens 1990 Visual scanning with or without spatial uncertainty and time-sharing performance. Unpublished paper, Aviation Research Laboratory, University of Illinois. Lockett, J.F., T. Plocher, and S. Dahl 1990 Crew reduction in armored vehicles ergonomic study (CRAVES). In E.J. Lovesey, ea., Contemporary Ergonomics. London: Taylor and Francis. Matthews, M.L. 1986 The influence of visual workload history on visual performance. Human Factors 28(6):623-632. Miller, G.A. 1956 The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review 63:81-97. Moray, N., E. Kruschelnicky, P. Eisen, L. Money, and I.B. Turksen 1988 The mental workload of skill-based, rule-based, and knowledge-based tasks and their combination. Pp. 164-167 in Proceedings of the Human Factors Society 32nd Annual Meeting. Santa Monica, California: Human Factors Society. Moray, N.P., and J. Liao 1988 A Quantitative Model of Excess Workload, Subjective Workload Estimation, and Performance Degradation. Report No. EPL-88-03. Urbana-Champaign: Univer- sity of Illinois, Department of Mechanical and Industrial Engineering. Mosier, K.L., and S.G. Hart 1986 Levels of information processing in a Fitts' Law task. Pp. 4.1-4.15 in Proceedings of the 21st Annual Conference on Manual Control. Report No. NASA CP-2428. Washington, DC: National Aeronautics and Space Administration. National Research Council 1990 Contact Lens Use Under Adverse Conditions: Applications in Military Aviation. Committee on Vision. Washington, DC: National Academy Press. Navon, D., and D. Gopher 1979 On the economy of the human-processing system. Psychological Review 86(3):214 255. Neisser, U. 1976 Cognition and Reality: Principles and Implications of Cognitive Psychology. San Francisco, California: Freeman.

WORKLOAD FACTORS 9 Norman, D.A., and D.G. Bobrow 1975 On data-limited and resource-limited processes. Journal of Cognitive Psychology 7(1):44-64. O'Donnell, R.D., and F.T. Eggemeier 1986 Workload assessment methodology. Pp. 42.1-42.49 in K. Boff, L. Kauffman, and J.P. Thomas, eds., Handbook of Perception and Human Performance. New York: Wiley and Sons. Pepitone, D.D., T. King, and M. Murphy 1988 The role of flight planning in aircrew decision performance. Proceedings of the SAE Aerospace Technology Conference and Exposition. SAE Technical Paper Series #881517. Warrendale, Pennsylvania: The Society of Automotive Engi- neers. Pepitone, D.D., R.J. Shively, and M.R. Bortolussi 1987 Pilot workload prediction. Proceedings of the SAE Aerospace Technology Confer- ence and Exposition. SAE Technical Paper Series #871771. Warrendale, Penn- sylvania: The Society of Automotive Engineers. Perrone, H.A. 1984 Visual slant misperception and the "black hole" landing situation. Aviation, Space, and Environmental Medicine 1020-1025. Postman, L., and L.W. Phillips 1965 Short-term temporal changes in free recall. Quarterly Journal of Experimental Psychology 17:132-138. Poulton, E.C. 1982 Influential companions: Effects of one strategy on another in the within-subjects designs of cognitive psychology. Psychological Bulletin 91:673-690. Prevost, M., and C.P. Banda 1991 Visualizing tool for human machine interface designers. Pp. 58-66 in Proceedings of the Society of Photo-optical Instrument Designers. Bellingham, Washington: Society of Photo-optical Instrument Designers. Rasmussen, J. 1982 The role of cognitive models of operators in the design, operation, and licensing of nuclear power plants. Pp. 13-35 in Proceedings of the Workshop on Cognitive Modeling of Nuclear Plant Control Room Operators. Report Nos. NUREG/CR- 3114; ONRL-TM-8614. Washington, DC: U.S. Nuclear Regulatory Commission. Rotondo, G. 1978 Workload and operational fatigue in helicopter pilots. Aviation Space and Envi- ronmental Medicine 430-436. Rouse, W.B. 1979 Approaches to mental workload. Pp. 255-262 in N. Moray, ea., Mental Workload: Its Theory and Measurement. New York: Plenum Press. Rouse, W.B., N.D. Geddes, and J.M. Hammer 1990 Computer-aided fighter pilots. IEEE Spectrum 27(3):38-40. Schneider, W., and R.M. Shiffrin 1977 Controlled and automatic human information processing: Decision research and attention. Psychological Review 84:1-66. Staveland, L.E., S.G. Hart, and Y.Y. Yeh 1986 Memory and subjective workload assessment. Pp. 7.1-7.13 in Proceedings of the 2Ist Annual Conference on Man7~al Control. Report No. NASA-CP-2428. Wash- ington, DC: National Aeronautics and Space Administration. Tanaka, K., A. Buharali, and T.B. Sheridan 1983 Mental workload in supervisory control of automated aircraft. Pp. 40-58 in Pro

92 WORKLOAD TRANSITION seedings of the 19th Annual Conference on Manual Control. Cambridge, Massa- chusetts: Massachusetts Institute of Technology. Thornton, D.C. 1985 An investigation of the "von Restorff' phenomenon in post-test workload ratings. Pp. 760-763 in Proceedings of the Human Factors Society 29th Annual Meeting. Santa Monica, California: Human Factors Society. Tsang, P.S., and C.D. Wickens 1988 The structural constraints and strategic control of resource allocation. Human Performance 1:45-72. Tulga, M.K., and T.B. Sheridan 1980 Dynamic decisions and workload in multitask supervisory control. IEEE Transac- tions on Systems, Man, and Cybernetics SMC-10:217-232. Vidulich, M., and M. Bortolussi 1988 A dissociation of objective and subjective workload measures in assessing the impact of speech controls in advanced helicopters. Pp. 1471-1476 in Proceedings of the Human Factors Society 32nd Annual Meeting. Santa Monica, California: Human Factors Society. Vidulich, M.A., and P.S. Tsang 1985 Techniques of subjective workload assessment: A comparison of two methodolo- gies. Pp. 239-246 in Proceedings of the Third Symposium on Aviation Psychol- ogy. Columbus, Ohio: Ohio State University. Vidulich. M.A., and C.D. Wickens 1984 Subjective workload assessment and voluntary control of effort in a tracking task. Pp. 57-72 in Proceedings of the 20th Annual Conference on Manual Control. Report No. NASA-CP-2341. Washington, DC: National Aeronautics and Space Administration. Causes of dissociation between subjective workload measures and performance. Applied Ergonomics 17 :291-296. Vienneau, R., and F. Gozzo 1987 Estimating pilots workload and its impact on system performance. Pp. 91-102 in Proceedings of the American Helicopter Society 43rd Annual Forum and Technol- ogy Display. New York: American Helicopter Society. Welford, A.T. 1967 A single channel operation in the brain. Acta Psychologica 27:5-22. Wickens, C.D. 1980 The structure of attentional resources. Pp. 239-257 in R. Nickerson, ea., Attention and Performance, Volume VIII. Englewood Cliffs, New Jersey: Erlbaum. 1986 The effects of control dynamics on performance. Pp. 39.1-39.60 in K.Boff, L. Kauffman, and J.P. Thomas, eds., Handbook of Perception and Human Perfor mance. New York: Wiley and Sons. 1989 Models of multitask situations. Pp. 259-273 in G. McMillan, ea., Applications of Models to System Design. New York: Plenum Press. 1991 Internal Model of Complex Dynamic Systems. Report No. ARI-RN-91-64. Urbana Champaign, Illinois: Illinois University at Urbana-Champaign Department of Psy chology. 1992a Computational models of human performance. Invited Address. Sixth Annual Conference on Psychology in the Department of Defense. U.S. Air Force Acad emy, Colorado. 1992b Engineering Psychology and Human Performance, 2nd Edition. New York: Harper Collins Publishers. 1986

WORKLOAD FACTORS 93 Wickens, C.D., I. Haskell, and K. Harte 1988 Ergonomic design for perspective flight-path displays. Proceedings of the 1988 International Conference on Systems, Man, and Cybernetics. New York: Institute of Electrical and Electronics Engineers. Wickens, C.D., and C. Kessel 1979 The effects of participatory mode and task workload on the detection of dynamic system failures. IEEE Transactions on Systems, Man, and Cybernetics SMC- 9(1):24 34 Wickens, C.D., D.L. Sandry, and M.A. Vidulich 1983 Compatibility and resource competition between modalities of input, central pro cessing, and output. Human Factors 25 (2) :227-248. Wiener, E. 1984 Vigilance and inspection. In J.S. Warm, ea., Sustained Attention in Human Per formance. London: Wiley. 1989 Reflections on human error: Matters of life and death. Pp. 1-7 in Proceedings of the Human Factors Society 33rd Annual Meeting. Santa Monica, California: Hu man Factors Society. Wiener, E.L., and R.E. Curry 1980 Flight-deck automation: Promises and problems. Ergonomics 23:995-1011. Woods, D.D. 1984 Visual momentum: A concept to improve the cognitive coupling of person and computer. International Journal of Man-Machine Studies 21(3):229-244. 1989 Modeling and predicting human error. Pp. 248-274 in J. Elkind, S. Card, J. Hochberg and B. Huey, eds., Human Performance Models for Computer-Aided Engineering. Washington, DC: National Academy Press. Yeh, Y.Y., and C.D. Wickens 1988 Dissociation of performance and subjective measures of workload. Human Fac- tors 30(1):111-120.

Next: 4 Stress »
Workload Transition: Implications for Individual and Team Performance Get This Book
×
Buy Paperback | $50.00
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

Workload transition is a potentially crucial problem in work situations wherein operators are faced with abrupt changes in task demands. People involved include military combat personnel, air-traffic controllers, medical personnel in emergency rooms, and long-distance drivers. They must be able to respond efficiently to sudden increases in workload imposed by a failure, crisis, or other, often unexpected, event.

This book provides a systematic evaluation of workload transition. It focuses on a broad spectrum of activities ranging from team cooperation to the maintenance of this problem on a theoretical level and offers several practical solutions.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!