You are on page 1of 14

MURI-OPUS Mental Resources and Performance Under Stress 1

DOD Multidisciplinary Research Program: MURI Operator Performance Under Stress (OPUS).

White Paper

On Mental Resources and Performance Under Stress

J.L. Szalma and P.A. Hancock

Department of Psychology and the Institute for Simulation and Training

University of Central Florida

Work Performed under the U.S. Army Research Office: MURI: Optimizing Cognitive Readiness Under Combat
Conditions

MURI-OPUS List of Collaborators

Universities: (Formal Connection): Kansas State University, University of Minnesota, Catholic University,
University of Cincinnati.
(Informal Connection): Florida Institute of Technology, University of Toronto.
Industry: United Defense, DCS Corporation.
Government: NAWC-TSD
Individuals: Dr. Jerry Krueger
MURI-OPUS Mental Resources and Performance Under Stress 2

Abstract

With the clear empirical failure of gated stage models of information processing to account for
multiple task performance outcomes, Kahneman (1973) offered resource theory as an alternative
explanatory construct. Considered primarily as a metaphor, resources have, over the years,
remained unidentified as a neurophysiological process or neuroanatomic structure. The failure
to establish resources as hypothetical constructs has been interpreted by some as the demise of
resource theory itself. Contemporary brain imaging techniques have begun to identify blood
flow, glucose transmission, and, potentially, correlates of cortical temperature as resource
foundations. The present paper elaborates upon resource theory, and articulates some
contemporary proposals on resource differentiation and resource utilization under stress as a
foundation for a program of theoretical and experimental progress.
MURI-OPUS Mental Resources and Performance Under Stress 3

ON MENTAL RESOURCES AND PERFORMANCE UNDER STRESS

Resource Metaphors

Most early resource theories, including Kahneman’s (1973) original view and modifications
by Norman and Bobrow (1975), Navon and Gopher (1979), and Wickens (1980; 1984),
conceptualized them as commodities or as pools of energy to be “spent” on task performance. In
general, either economic or thermodynamic/hydraulic metaphors tend to be adopted when
describing resources. The economic model is reflected in the description of resources in terms of
supply and demand: Performance on one or more tasks suffers when the resource demands of
the tasks exceed available supply. Presumably, the total amount of this supply fluctuates with the
state of the individual, with the “assets” diminishing with increases in the intensity or duration of
stress. Although Kahneman’s (1973) original conception allow for dynamic variation in resources,
most early models assumed a fixed amount of resources (see Navon & Gopher, 1979). In
thermodynamic analogies, resources are a fuel that is consumed, or a tank of liquid to be divided
among several tasks, and under stressful conditions the amount of resources available is depleted
and performance suffers. In discussing his version of resource theory, Wickens (1984) warned
that the hydraulic metaphor should not be taken too literally, but most subsequent descriptions
of resources have employed visual representations of resources as just this tank of liquid form.
Similarly, many discussions of resource availability and expenditure adopt the economic
language of supply and demand, and Navon and Gopher (1979) explicitly adopted principles of
microeconomics in developing their approach.

A problem with using non-biological metaphors to represent biological systems is that such
models often fail to capture the complexity and the unique dynamic characteristics (e.g. adaptive
responses) of living systems. For instance, a hydraulic model of resources links the activity of a
tank of liquid, governed by thermodynamic principles, to the action of arousal mechanisms or
energy reserves that are allocated for task performance. However, a thermodynamic description
of the physiological processes underlying resources is at a level of explanation that may not
adequately describe the psychological processes that govern performance. Thermodynamic
principles can be applied to the chemical processes that occur within and between neurons, but
they may be less useful in describing the behavior of large networks of neurons 1. Similarly,
economic metaphors of supply and demand may not adequately capture the relation between
cognitive architecture and energy allocated for their function. Economic models of resources
define them as commodities to be “spent” on one or more activities, and they assume an
isomorphism between human cognitive activity and economic activity, an assumption which
may not be tenable. Indeed, Navon and Gopher (1979) admitted that their ‘static’ economic
metaphor for multiple resources may need to be replaced by a dynamic one that includes
temporal factors (e.g. serial versus parallel processing; activity of one processing unit being
contingent upon the output of another). Such concerns over the metaphors used to describe
resources are hardly new (Navon, 1984; Wickens, 1984), but their use has become sufficiently
ingrained in thinking about resources and human performance that reevaluation of the
metaphors is warranted. As we shall see, a regulatory model based on physiology may serve as a

1
The argument here is not that neural structures are not constrained by the laws of thermodynamics—
clearly they are—but that thermodynamic principles implied by the metaphor are not sufficient for the
development of a complete description of resources and their relation to cognitive activity.
MURI-OPUS Mental Resources and Performance Under Stress 4

better metaphor (and, in future, may serve to describe resources themselves to the extent that
they can be established as a hypothetical construct) to describe the role of resources in human
cognition and performance. However, even a physiologically-based theory of resources must be
tempered by the problems inherent in reducing psychological processes to physiological activity.

Development of a Single Resource Theory

As a precursor to the emergence of resource theory, Moray (1967) argued that Broadbent’s
(1958) filter theory, and Treisman’s (1969) revision, does not apply to all tasks requiring divided
or selective attention. Specifically, these models do not adequately account for the performance
of well-trained observers on divided attention tasks, or for cases in which the perceptual inputs
and the responses (outputs) are compatible. Using the language of Information Theory, he
argued that the human should be considered a limited capacity processor rather than a fixed,
limited capacity channel. It is the limitation of the central processor that influences performance
rather than the perceptual input channels. The mental operations performed on perceptual
inputs consume capacity (an aspect of attention not considered by Broadbent or Treisman), and
capacity can be divided among different “processors” when divided attention is required. Moray
was describing what would later be called “allocation of resources.” He also argued that the
demands of the task control how this allocation occurs (the capacity allocated to a given channel).
Note that in this context capacity is viewed in terms of information processing structures or
channels rather than as pools of energy posited by more energetic based models.

An early influential resource theory was that of Kahneman (1973), who argued that there was
a single pool of limited capacity available for a variety of tasks, and that performance depends
upon the degree to which that capacity is allocated to the task. The amount of capacity available
was assumed to be limited and a function of arousal level. An allocation strategy determines
how much processing capacity various “activities” (presumably this includes cognitive
processes) receive, and the strategy adopted is influenced by characteristics of the individual
(“enduring dispositions”) and motivational influences (“momentary intentions”).

Kahneman’s (1973) theory was extended by Norman and Bobrow (1975), who are credited
with coining the term “resources” in the context of attention and human performance. They
distinguished resource-limited from data-limited processes. Resource-limited processes are those
that improve as resource investment is increased, and data-limited processes are those in which
no further increase in resource allocation will impact performance, which is driven by the quality
of the information (data) available to the individual. According to Norman and Bobrow, there
are cases in which performance is resource independent; this has implications for optimal
resource allocation. The effectiveness of allocation depends not only on environmental demands
but also on limitations imposed by the structure of the task. The distinction between resource
and data-limited processes can account for “flat” portions of the performance-resource function,
in which performance is unaffected by increases in resource allocation or effort. Note, however,
that in some situations an apparent data-limited process may actually be a simple ceiling effect
created by the level of task difficulty chosen by the experimenter (Kantowitz & Knight, 1976),
although such a conclusion is somewhat controversial (see Norman & Bobrow, 1976).

The data-limited process concept was expanded by Navon and Gopher (1979) who used the
term subject-task parameters, which include the task characteristics (quality of the data), the
environment, and the permanent properties and transient states of the subject. Hence, their view
MURI-OPUS Mental Resources and Performance Under Stress 5

extended data-limited processes to include task-operator interactions, a conceptualization


consistent with transactional models of stress (Lazarus & Folkman, 1984; Matthews, 2001). The
subject-task parameters define the constraints within which resources can be recruited and
allocated for task performance. In addition, they noted that rather than being a fixed capacity
system, resources may be somewhat flexible, fluctuating with arousal, such that in some
situations increasing task load may increase arousal and release a larger supply of resources.

Multiple Resources

An alternative to a single resource view is multiple resource theory. Navon and Gopher
(1979) argued that the human is a multiple channel processor (multiple structures), that each
processor may have its own capacities, and that each capacity may be shared by several
processes. Hence, the impact of changes in task difficulty on resources may not be purely
quantitative, but may also be qualitative (i.e. structural). Different subject-task parameters may
differentially influence the resource demands on diverse processing structures, and specific
parameters may influence the amount of resources demanded for a particular structure. A
question to be addressed is whether different stressors differentially affect specific information
processing structures available, or whether the effects of stressors are more global in character
(i.e. affecting all structures simultaneously, either to the same or to different degrees). Navon
and Gopher only mentioned stress in the context of arousal and resources, indicating that high
arousal accompanying high task load is a state of stress rather than of processing capacity.
However, this brief treatment implies an arousal model of stress without considering the multi-
dimensionality of the stress construct or the impact of stress on cognitive structures and the
quality of information processing (see Matthews, 2001).

In an influential version of multiple resource theory, Wickens (1984) proposed separate pools
of resources for spatial and verbal information processing codes, different modalities, 2 stages of
information processing (encoding, central processing, and responding), and motor vs. verbal
response. These resources can be utilized independently or jointly depending on the demands of
the information processing components that comprise a task.

Multiple resource views developed as a result of experimental findings not easily accounted
for by single resource views. For instance, Wickens (1984) argued that single-resource theories
cannot readily explain perfect time-sharing between tasks, nor can they explain why changes in
task structure that leave difficulty constant (and, presumably, the amount of resources required
for task performance) lead to a change in the degree of interference with a secondary task. In
addition, Wickens described two aspects of task difficulty that cannot easily be accounted for by
single resource views: 1) Difficulty insensitivity; increases in the demands imposed by a primary
task do not influence performance on a secondary task. 2) Uncoupling of difficulty and structure;
cases in which two tasks are each paired with a third task, but the more difficult of the two tasks
interferes less with the third task than the easier one.

Wickens argued that a multiple resource model can account for each of these issues. Perfect
time sharing can occur when there is no overlap in resource demand between two tasks, as

2
The original model specified only audition and vision, but in principle the model would apply to other
modalities as well.
MURI-OPUS Mental Resources and Performance Under Stress 6

would be the case with an auditory shadowing task and a secondary tracking task. Cases in
which task structure changes do not influence the overall amount of resources required (shifting
from an auditory to a visual task) but impair secondary task performance is due to a shift in
resource demands from separate pools to a common reservoir of resources. Difficulty
insensitivity is accounted for by cases where the two tasks impose demands on different pools of
resources, and uncoupling of difficulty and structure occurs when the more difficult task shares
resources with the third task but the easier task does not. In arguing for the existence of multiple
resources, Wickens (1984) was careful to point out that this view does not eliminate the
possibility of an additional set of general resources available for use in any task.

Definition of Resources

A major problem for resource theory, whether single or multiple, is defining the concept
itself. Wickens (1984) defined resources as an intervening variable to account for variability in
the efficiency with which individuals can divide attention among tasks. He made the distinction
between resources as something available for use (a “commodity”), versus resources as distinct
information processing structures, in which the degree to which two tasks interfere with one
another is determined by the degree to which they compete for common structures (e.g. spatial,
verbal). He also argued (Wickens, 1992) that capacity, which represents the maximum or upper
limit of processing capability, should not be confused with resources, which refer to scarce
supplies of mental effort allocated to improve processing efficiency. Alternatively, Gopher (1986)
argued that structure and energy (or the “commodity” in Wickens’ terms) are distinct
components of resources. Indeed, the multiple resources concept inherently includes a structural
element. However, this distinction is blurred by ambiguous terms like “processing resources,”
which can be interpreted as energy or structure (or both). This has contributed to the confusion
surrounding the concept.

There have been attempts to identify physical structures that serve as resources. Wickens
(1984) discussed several candidates, which if successful would allow the concept of resources to
shift from its status as an intervening variable to that of a hypothetical construct. He suggested
that pupil diameter may be a useful index of resources, as it is a manifestation of ARAS activity.
However, such a view emphasizes resources as energy reserves and neglects the importance of
structure inherent in a multiple resource view. To support the latter view, Wickens noted that
there is some evidence for “functional cerebral space,” for which different tasks compete and
which may ultimately serve to define resources as neural processing mechanisms (see
Kinsbourne & Hicks, 1980). From this perspective, there is resource competition to the degree
that task demands share the same neural structures. This implies that resources are processing
capacity in the computer sense, where the processing chip and working memory of the computer
are used as a metaphor for psychological processing (a third metaphor applied to resource theory
more recently, although see Moray, 1967).

It is also possible that the cognitive structures (information processors) are the resources.
Rugg (1986) suggested that the need for the concept of resources arises from the current inability
to adequately specify the cognitive architecture responsible for information processing. If the
specific components of information processing structures that drive performance could be
articulated, then the number of resources would be the same as the number of active
components; the resource model thus becomes unnecessary (see also Navon, 1984). Resource
MURI-OPUS Mental Resources and Performance Under Stress 7

limitations reduce to structural limitations when the structures are adequately specified.
However, such a view, while acknowledging the importance of cognitive structure to resource
theory, does not reflect the importance of the energetic component of behavior. In addition,
Rugg himself argued that the link between resources and physiology was not clear, since the
physiological changes could be a consequence of the mobilization of resources (the body’s
attempt to adapt to the behavioral demands).

Resources: Structure and Energy. As noted earlier, Gopher (1986) argued that resources
consist of structural and energetic components. Specifically, resources are aggregations of
elementary units defined by the type of processing activity they serve and the range and rate of
activation. The degree of activation of those units is determined by the energetic component.
Gopher noted that most views of resources emphasized the energetic over the structural
elements, an observation that is true sixteen years later. He acknowledged, however, that
Kahneman’s (1973) original model did include structural components in the form of the possible
activities to which available capacity (resources) was allocated. Wicken’s multiple resource
model also contains structure, inherent in the definitions of the qualitatively different pools of
resources (e.g. stages of processing, coding, etc). Gopher’s view of structure, however, goes
beyond these general conceptions to consider specific units of information processors that
comprise, in his model, a vector for structure that combines with energetic sources for each
processor, the product being system response. Hence, in Gopher’s model, the response of the
overall information processing system is a product of two vectors: structure and energy.

If resources consist of both cognitive structure and energy, then the concepts of resource
investment and scarcity of resources have multiple interpretations, which has likely contributed
to confusion in the development and tests of the theory. Gopher (1986) addressed the different
manifestations of both resource investment and scarcity. He described three forms of modulation
in intensity of resource investment: 1) changing the number of processing units engaged in a
task, 2) changing the activation level of those units, or 3) changing the time each processing unit
spends on a task. The first and second forms of resource investment are the most often employed
in applications of resource theory to research on stress and performance, a bias that may be due
in part to the history of the theory as an answer to the problems of arousal models in explaining
the effects of stress on performance. This has led to the neglect of the relation of the third form of
investment to stress (although see Hancock and Weaver, in press, for a noteworthy exception).

While scarcity of resources can be described in terms of the amount available for
consumption, Gopher (1986) discussed another aspect of scarcity that is consistent with a
physiological analogy: the rate at which they can utilize resources. This represents the structural
limitations of the processing units. For instance, if a car has a full tank of gas, but the flow of fuel
to the engine is maximized, no increase in the pool of resources will help, even though
performance appears “resource limited.” Here the scarcity in resources is not in their amount,
but in their rate of usability. He also noted that time and whether resource scarcity is transient or
steady state can also influence the nature of resource limitations. Note that this view of
resources and its implications for investment and scarcity is beyond the relatively simplistic
economic and thermodynamic metaphors often applied in descriptions of the theory.
MURI-OPUS Mental Resources and Performance Under Stress 8

A Physiological Model

Recently, Cabanac and Russek (2000) outlined a representation of regulated biological


systems that may be useful for conceptualizing the energy component of resource theory. They
adopted characteristics of control theory and applied them to biological systems. Physiological
systems reach a steady state, such that a constant amount of free energy available for use is
maintained and the inflow and outflow of energy are equal and constant. These systems have a
set point that indicates the normal level of function. Perturbations of this steady state require the
system to compensate for deviations from the set point (this is essentially a homeostatic model).
Further, Cabanac and Russek (2000) argued that the regulation of inflow is anatomically distinct
from the regulation of outflow. The inflow regulation is negative feedback loop in which a
standard reference (presumably determined by the set point) is compared to current state and an
input subsystem responds by increasing flow when energy is drained and decreasing it when
energy levels rise above the set point (a negative loop). The outflow is a positive (feedforward)
loop in which changes in the state relative to the set point induce changes in the same direction in
the output subsystem (e.g. increase in amount of energy induces increase in outflow).

They also distinguished between regulated variables, those that are maintained at a constant
level, and controlled variables, which correct the deviations of the regulated variables. They
delineated three properties of any system which continuously modulates the inflow/outflow of
energy or mass: 1) the minimum flow rate through the system, which is the set point for the
regulated variable. Perturbations change either inflow or outflow, which changes the total flow
rate. 2) The maximum continuous perturbation the system can regulate against, which is equal to
the maximum flow rate of the system [this presumably is limited by the structure of the system—
e.g. the neuro architecture and function limitations; at higher levels, the processing capacity of
attention and cognition]. 3) The speed at which the system can return to set point after transient
perturbation is proportional to the magnitude of the baseline flow rate.

Although Cabanac and Russek (2000) argued that inflow and outflow processes were
anatomically distinct, it may be difficult to empirically distinguish them at the level of perception
and cognition (and they may not be anatomically distinct when applied to the nervous system).
If one applies their model to psychological processes, the “tank” of energy becomes a pool of
resources (as energy) which are continuously drained and replenished, thus creating a steady
state. Perturbations would result in deviations from set point (steady state) and require
regulatory response in the form of either increasing/decreasing inflow or outflow, and the
depletion of resources would occur when outflow is greater than inflow for a specified period of
time. The regulatory response would presumably be controlled by the cognitive structures
responsible for resource allocation (e.g. a “central executive;” see Baddeley, 2001). Stressors, as a
perturbation, require increasing the inflow, since outflow cannot be decreased without sacrificing
task performance. Assuming a motivated operator, performance declines as the ability to
replenish diminishes. Alternatively, stress could increase outflow, as more processors are
engaged and more energy expended to cope with the stress while maintaining performance. In
either case, a threshold (for the amount of available resources) may exist below which
performance begins to suffer. Note that an implication of the feeedforward loop is that one can
have “too many” resources, such that the amount of resources available exceeds steady state
levels. In such situations, a person would seek activities to consume those resources and
performance might suffer as a result of understimulation (see Hancock & Warm, 1989).
MURI-OPUS Mental Resources and Performance Under Stress 9

This model addresses the energetic component of resources, but it does not address
structural considerations, beyond the maximum perturbation a system can handle or how
resources are allocated for information processing, both of which are presumably constrained by
structural factors. The maximum amount of energy that can be applied to compensate for
perturbation will, if we apply Cabanac and Russek’s (2000) model, be equal to the maximum flow
rate of the system, presumably a structural feature. This may represent a point at which a
process becomes data-limited. The characteristics of the information processors, however,
remain unspecified by the physiological model. An additional structural concern is whether the
quality of the information processing degrades (independent of energy level) with increases in
the intensity and/or duration of stress.

Issues Surrounding the Resource Concept

There are several issues regarding resource theory that need to be addressed if it is to be
useful for advancing stress theory and clarifying the link between stress and performance. These
concerns are addressed in this section.

Single vs. Multiple Resources. It is possible that rather than a set of qualitatively different
resources, there may be a single resource that can be more easily (“naturally”) divided for
particular sets of activities. An analogy from classical conditioning may clarify the issue. Certain
S-R relations are learned faster than others. For instance, learning to avoid a stimulus after it is
associated with nausea can occur in one trial, while learning to avoid the same stimulus when it
is paired with shock is learned more slowly (Rozin & Kalat, 1971). Similarly, it may be that
processing resources are more easily separable for some sets of tasks than for others. Indeed,
Moray (1967) noted that selection of attention is easier along some dimensions than others.
Hence, the differences in time sharing performance may have to do with ease of allocation of a
single resource rather than multiple resources per se. However, these two concepts may be
empirically indistinguishable. Ultimately, the issue reduces to the more fundamental problem
regarding the identity of resources, as discussed later in this section.

Function of Resources. Another problem for resource theory is the absence of a precise
description of how resources control different forms of information processing. Do resources
determine the energy allocated to an information processor, do they provide the space within
which the processing structure works, or does the processor draw on the resources as needed
(and available). In the latter case, the cognitive architecture would drive energy consumption
and allocation, but the locus of control for the division of resources remains unspecified in any
case. Presumably, an “executive” function that either co-ordinates information processors
drawing on different pools of resources or decides how resources will be allocated must itself
consume resources, in terms of both energy required for decision making and mental “space” or
structure required. Hence, resource theory does not solve the homunculus problem for theories
of attention, nor does it adequately describe resource allocation strategies behind performance of
information processing tasks.

Empirical Tests of the Model. Navon and Gopher (1979) commented on the problem of
empirically distinguishing declines in performance due to insufficient supply from those
resulting from increases in demand. They noted that, “When the performance of a task
deteriorates, is it because the task now gets fewer resources or because it now requires more?” (p
243) This is similar to the issue concerning whether resource depletion results from changes of
MURI-OPUS Mental Resources and Performance Under Stress 10

inflow or outflow discussed earlier in the context of Cabanac and Russek’s (2000) physiological
model. Navon and Gopher (1979) characterized the problem as distinguishing between changes
in resources and changes in the subject-task parameters that constrain resource utilization, and
they offered two approaches to avoid this difficulty. One approach is to define the fixed
constraints of the task and observe how the information processing system manages the
processes within those constraints. The degree of freedom of the system, in this view, is the pool
of resources available, in which the term ‘resource’ is interpreted broadly to include quality of
information, number of extracted features, or visual resolution. The subject-task parameters
define what is imposed on the system (the demands) and the resources refer to what the system
does in response to the demands (allocation of processing units). From this perspective resources
can be manipulated by the information processing system within the constraints set by the
subject-task parameters. A second approach is to distinguish the kind of control the system exerts
on resources, between control on the use of processing devices (what we have called “structure”)
and the control of the properties of the inputs that go into these devices. The ‘devices’ are
processing resources. The other kind of control is exerted on input resources, which represents
the flexibility the person has for determining which inputs are operated on, as determined by
subject-task parameters. Processing resources are limited by the capacities of the information
processors, while the input resources are limited by subject-task parameters (and allocation
strategies that determine which information the operator attends to). Presumably, the individual
would have some control over the allocation strategy, in terms of the processing resources
devoted to a task, although these can also be driven by task demands (e.g. a spatial task requires
spatial processing units). Navon and Gopher did not advocate either approach, but presented
them as alternatives for further investigation. The implication for examining the resource model
of stress is that one must manipulate both the subject-task parameters (e.g. by varying the
psychophysical properties of the stimulus, manipulating the state of the observer, or varying the
kind of information processing demanded by the task) as well as the allocation strategies the
operator uses (the input resources—e.g. payoff matrices, task instructions). This would provide
information regarding how specific stressors impair specific information processing units and
how they change operator resource allocation strategies in the presence of continuous stress.

In a later article, Navon (1984) moved to a position less favorable toward resources than the
earlier approach (Navon & Gopher, 1979). In the later paper, Navon asserted that predictions
derived by resource theory could be made, and results explained, without appealing to the
resource concept (see also Rugg, 1986). One could instead interpret effects in terms of the
outputs of information processors. Most manipulations, such as difficulty (which in his view
influences the efficiency of a unit of “resources”) or complexity (which affects the load, or the
number of operations required) influence the demand for processing, with supply having no
impact upon their interaction. However, this approach assumes a clear distinction between
outputs of a processing system and the concept of a resource, and Navon’s (1984) notion of
specific processors seems blurred with the notion of a resource, as both are utilized for task
performance. Nevertheless, his critique that the resource concept is too vague is relevant, and
Navon did argue that if resources are viewed as an intervening variable rather than a
hypothetical construct, the concept has utility for describing the process. According to Navon, it
doesn’t matter whether the economic terminology is meant literally or metaphorically, although,
as noted earlier, metaphors taken too literally can be misleading.

Structural Mechanisms. If different kinds of information processing draw on different kinds


of resources, in terms of the information processors engaged in a task, stressors may have
MURI-OPUS Mental Resources and Performance Under Stress 11

characteristic effects on each “resource.” In addition, as Navon and Gopher (1979) have noted, an
aspect of resource utilization is the efficiency of each resource unit. It may be that stress degrades
the efficiency of information processing units, independent of energy level or allocation strategy.
Investigation of such effects could be accomplished by transitioning between tasks requiring
different kinds of information processing and determining if the effects of stress on one structure
impacts the efficiency of a second structure.

The quality of resources can vary not only in terms of the kind of information processing unit
engaged, but also in terms of the kind of task required. Following Rasmussen’s (1983)
classification system for behavior, some tasks require knowledge-based processing, in which the
operator must consciously rely on his/her mental model of the system in order to achieve
successful performance. Other tasks fall under the category of rule-based behavior, in which a
set rules or procedures define task performance. The third category is skill-based behavior, in
which the task is performed with a high degree of automaticity. Presumably, each kind of task
requires different amounts of resources, but they may also represent qualitatively different forms
of resource utilization. In other words, these tasks may differ in the efficiency of a unit of
resources. As task performance moves from knowledge to rule to skill based processing (e.g.
with training), the cognitive architecture may change such that fewer energy reserves are
required, and those that are engaged become more efficient. Moreover, the way in which each of
these systems degrade with time under stress may be systematic, with the more fragile
knowledge-based processing degrading first, followed by rule based processing, with skill based
processing degrading last (at this point, one may begin to see breakdown of not only
psychological processes but physiological ones as well—see Hancock & Warm, 1989). This
degradation may follow a hysteresis function, such that a precipitous decline in performance
occurs as the operator’s resource capacity is reduced below a minimum threshold for
performance. Moreover, these processes may recover in an inverse form, with skill-based
processing recovering first, followed by rule and knowledge-based processing. However, based
upon the physiological model outlined earlier, one would expect that expanding available
resources (opening up the inflow valve) would allow more resources to be applied to tasks and
would have a greater impact upon knowledge rather than rule or skill based processing.

Energetic Mechanisms. To investigate the energetic aspects of resources, one must


manipulate environmentally-based perturbations, in the form of external stressors (noise, heat)
and task demands, to systematically affect inflow versus outflow of energy. Presumably, inflow
is controlled by arousal levels, physiological energy reserves, and effort. One could examine
performance under manipulations of energetic resources under dual task performance (e.g. what
happens to performance on two tasks under sleep deprivation or caffeine consumption?). For
example, the steady state can be perturbed by increasing (e.g. caffeine) or decreasing (e.g. sleep
deprivation) energy while systematically varying the demands for two tasks.

Time is an important component of the process of resource recruitment and allocation, as


changes in mental resources occur as a function of time on task. Thus, transitions from one task
demand level to another could be investigated in a systematic fashion under conditions of
high/low stress. The key challenge is determining whether stressors such as noise, heat, or sleep
deprivation slows inflow, increases outflow, or whether they have interactive effects. We must
assume that outflow is manifested by behavior at multiple levels (from cellular up to molar
behaviors and cognitive processing). Hence, to increase outflow, one could manipulate task
demands, in terms of complexity or difficulty. The Hancock and Warm (1989) model applies to
MURI-OPUS Mental Resources and Performance Under Stress 12

this situation, as increasing effort to cope with stress progressively impairs psychological and
eventually physiological functioning. The interaction between inflow and outflow is manifested
in the interaction between effort and co-ordination of task demands. It is an open question
whether the allocation of resources occurs at the outflow point (multiple spigots), or at the input
(multiple resources fed by a central pool), an issue that may not be easily amenable to empirical
testing.

Structure & Energy. Another empirical challenge is to distinguish structure from energy.
Given the definitional problems associated with the resource concept, it is not clear whether
performance declines because of reduction in energy level or degradation in structures (i.e.
failures or declines in the efficiency of the processing units), or a combination of both. If structure
and energy are distinct elements of resources, it is hypothetically possible to manipulate one
while holding the other constant, although the validity of that assumption is questionable. Is it
possible to manipulate specific forms of information processing under constant energy level? Is it
possible to manipulate energy level independent of which cognitive processes are utilized? If the
decline in available resources is, at least in part, due to the degradation of particular information
processing units, then transferring to a task requiring the same processor should lead to worse
performance than transferring to one that is different. For instance, if a person engages in a task
requiring verbal working memory while under stress, then transition to a task requiring spatial
discrimination, performance on the latter should depend only on energetic factors, not on
structural ones. Note, however, that in this case the effects of different mental capacities would be
confounded with the effects of novelty and motivation on performance.

Summary

Resource theory was developed as a response to the limitations in the ability of alternate
information processing models, such as bottleneck theories of attention and the arousal model of
performance, to account for results on performance of tasks requiring selective and divided
attention. Resources are intervening variables thought to control human performance and
operator response to workload and stress. Some have argued that resources may be a
hypothetical construct, but strong evidence for this has not been forthcoming, mainly because it
is difficult to establish a single biological structure or set of structures that correspond to
resources. Both economic and energetic metaphors have been used in describing the structure
and function of mental resources. This review of resource theory described resource models of
attention and human performance, and modifications of those models. The limitations of
resource theory were addressed, and the implications of both the theory and its limitations for
stress theory were discussed. Specific research questions based upon these concerns are listed
below.

Identified Research Questions

• Can a regulatory model based upon physiological processes be applied as a metaphor for
resources? Is the energetic element of resources controlled by such a mechanism?

• Can the structural component of resources (information processing units) be specified


beyond the currently gross level?
MURI-OPUS Mental Resources and Performance Under Stress 13

• In addition to resource depletion, there may be strategy changes that occur with time on
task. This relates to the differential quality of resources.
o Do knowledge, rule, and skill based processing degrade progressively?
o After a stressor is removed, do these three processes recover in reverse order?

• Are the processes of depletion and replenishment of resources empirically


distinguishable?
o If so, is resource scarcity a consequence of depletion over time, a slowing in
replenishment processes over time, or an interaction?
o If energy reserves are manipulated (caffeine, fatigue) and task demands are
manipulated (drawing more resources), can relative rates of “inflow” and
“outflow” of resources be explored?

• Are there thresholds for resource depletion?


o If so, do these thresholds vary according to cognitive architecture (the specific
forms of information processing employed for a task)?

• Does threshold level relate to hysteresis, such that the precipitous decline in performance
occurs as an operator’s resources decrease below threshold, and recovery occurs as
resources are replenished to threshold levels?

• If resources consist of structural and energetic components, how does stress impact each
one? How does stress influence the interaction of these components?

• Do factors that degrade energy levels, such as sleep deprivation or temperature


fluctuations, impact knowledge based processing more severely than the other forms of
processing?

Anticipated Outcomes

• Increased understanding of how stress impacts information processing and mental


resource utilization will promote a deeper understanding of the mechanisms by which
stress impacts performance.

• Increased understanding of the relation between resources and stress will facilitate
development of theoretical models of stress and performance.

• Identification of factors (task, environment, person) that control resource allocation and
consumption.

• A description of the implications of advances in resource theory and research for the
design of stress-adapted human-machine interfaces.
MURI-OPUS Mental Resources and Performance Under Stress 14

References
Baddeley, A.D. (2001). Is working memory still working? American Psychologist, 56, 851-864.

Broadbent, D.E. (1958). Perception and communication. London: Pergamon Press.

Cabanac, M. & Russek, M. (2000). Regulated biological systems. Journal of Biological Systems, 8, 141-149.

Gopher, D. (1986). In defense of resources: On structures, energies, pools, and the allocation of attention. In G.R.J
Hockey, A.W.K. Gaillard, & M.G.H. Coles (Eds.), Energetics and human information processing. (pp. 353-371).
Dordrecht: Martinus Nijhoff Publishers.

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

Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ: Prentice Hall.

Kantowitz, B.H., & Knight, J.L. (1976). On experimenter-limited processes. Psychological Review, 6, 502-507.

Kinsbourne, M., & Hicks, R.E. (1980). Functional cerebral space: A model for outflow, transfer, and interference effects in
human performance: A tutorial review. In R. Nickerson (Ed.), Attention and performance VIII. (pp. 345-362). Hillsdale,
NJ: Erlbaum.

Lazarus, R., & Folkman, S. (1984). Stress, appraisal, and coping. New York: Springer Verlag.

Matthews, G. (2001). Levels of transaction: A cognitive science framework for operator stress. In P.A. Hancock & P.A.
Desmond (Eds.), Stress, workload, and fatigue. Mahwah, NJ: Erlbaum.

Moray, N. (1967). Where is capacity limited? A survey and a model. Acta Psychologica, 27, 84-92.

Navon, D. (1984). Resources—A theoretical soupstone? Psychological Review, 91, 216-234.

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

Norman, D., & Bobrow, D. (1975). On data-limited and resource-limited processing. Journal of Cognitive Psychology, 7,
44-60.

Norman, D. & Bobrow, D. (1976). On the analysis of performance operating characteristics. Psychological Review, 6,
508-510.

Rasmussen, J. (1983). Skills, rules, and knowledge; Signals, signs, and other distinctions in human performance models.
IEEE Transactions on Systems, Man, and Cybernetics, vol. SMC-13, 257-266.

Rozin, P., & Kalat, J.W. (1971). Specific hungers and poison avoidance as adaptive specializations of learning.
Psychological Review, 78, 459-486.

Rugg, M.D. (1986). Constraints on cognitive performance: Some problems with and alternatives to resource theory. In
G.R.J Hockey, A.W.K. Gaillard, & M.G.H. Coles (Eds.), Energetics and human information processing. (pp. 353-371).
Dordrecht: Martinus Nijhoff Publishers.

Treisman, A.M. (1969). Strategies and models of selective attention. Psychological Review, 76, 282-292.

Wickens, C.D. (1980). The structure of attentional resources. In R. Nickerson (Ed.), Attention and performance VIII.
(pp.239-257). Hillsdale, NJ: Erlbaum.

Wickens, C.D. (1984). Processing resources in attention. In R. Parasuraman & D.R. Davies (Eds.), Varieties of attention.
(pp. 63-102). New York, NY: Academic Press.

Wickens, C.D. (1992). Engineering psychology and human performance, 2nd Ed. New York, NY: Harper Collins.

You might also like