1- Research Proposals
1
Chapter Research Brief
Informing Science, Volume 1, Chapter 9 – Receivers and Clients
In this chapter, the author, Grandon Gill transitions from the previous discussion (Chapter 8) of
the role of Senders and Informers in an information system to the describing the challenges
encountered by the client/receivers.
Though the bulk of the chapter focuses on the challenges and biases encountered by individual
clients, Gill interestingly begins with a historical vignette about the organizational (i.e. networks and
multiple clients) resistance encountered by a naval officer, William Sims, in his efforts to inform the
Department of the Navy about a new and better method of executing maritime gunfire. The
introduction of the “Case Study: Morison’s Gunfire at Sea” provided a useful method as many of the
challenges described in the case were resultant from a variety of individual biases that Gill then goes on
to describe.
Gill provides a basic model to detail the various challenges that client’s may have to fully and
accurately receiving intended information. He introduces a “Single Client Resonance Model” which may
help explain why informing messages can be distorted (i.e. stopped, altered, distorted) by the recipient.
The author then walks through and explains the specifics of the various client filters. In explaining the
overall model and the detailed characteristics of his bias filter model, Gill cites various authorities and
studies to substantiate and/or exemplify the client filters, including Attention filters, Information filters,
Cognitive filters, Risk and Time preference, Motivation filters and Visceral filters. It was also noted that
in addition to describing client visceral filers, Gill dedicates substantial discussion to the impact that
these particular filters have on cognition and on choice processes. Further, he also describes the
impacts of emotions and social factors on establishing the visceral filter In general, the bias model and
detailed description of its components was well articulated, easy to follow and supported by relevant
research. That said, there were also some clear shortfalls in some of the supporting research
substantiation.
Following the single client bias model discussion, Gill transitions to a discussion of the challenges
of multiple clients. No real discussion is provided by Gill beyond presenting an excerpt by Andreas
Drechsler, “Designing to Inform: Toward Conceptualizing Practitioner Audiences for Socio-Technical
Artifacts in Design Science Research in the Information Systems Discipline.” The difficulties and
challenges associated with informing networks or multiple clients was not discussed and made the
except seem ill placed within this chapter. Lastly, the chapter closed with the Innerscope Case Study.
This was an interesting case that focuses on current efforts in the biometric and neuroscience fields in
order to mitigate several informing challenges. This case study was useful insofar as it was current and
relevant with respect to how commercial companies are using technological advances to measure
physiological changes to potential determine future behavior.
The bottom line is that this chapter was informative in establishing the importance of
and factors surrounding client biases in the informing process. The latter half of the chapter however,
did not seem to logically follow and seemed out of place.
2
Citation Credibility Fidelity
Jamieson, K., & Hyland, P. (2006). Good intuition or fear and uncertainty: The effects of bias on information systems selection decisions. Informing Science: The International Journal of an Emerging Transdiscipline, 9, 49-69.
Published in a credible academic journal -Informing Science – and edited by Eli Cohen, a preeminent scholar in the field. The authors - Jamieson, a Phd candidate – and Hyland, professor at Queensland University in Australia. Hyland has numerous publications and citations to his credits.
Gill uses the cited model as a launching point or point of departure for his augmented model. Citing this was appropriate to indicate that the underlying picture wasn’t developed from scratch. Gill accurately captures and appropriately applies what Jamieson and Hyland presented.
Driver, M. J., & Streufert, S. (1969). Integrative complexity: An approach to individuals and groups as information-processing systems. Administrative Science Quarterly, 272-285.
Both authors are well established in their fields. Driver is a respected author researcher having numerous publications to his credit. Many of his works having been translated to foreign journals. Michael J. Driver is associate professor of management at the University of Southern Cali- fornia. Siegfried Streufert is associate professor of psychology at Purdue University. This article was cited almost 100 times in other works. That said, while the research may still be valid, it is dated.
Gill uses this citation to support the claim that “Although we exhibit a natural curiosity, which is to say we have a cognitive motivation for information seeking” However, the cited work focuses on information processing systems and does not make any reference to natural curiosity. Use of this citation did not seem appropriate.
Hackman & Oldham 1980
Gill cites Hackman and Oldham within the chapter, but there is no accompanying information in the reference section.
Unkown whether use of the cited work supports what Gill states.
3
Loewenstein, G. (1996). Out of control: Visceral influences on behavior. Organizational behavior and human decision processes, 65(3), 272-292.
1207 citations. George Freud Loewenstein is an American educator and economist. He is the Herbert A. Simon Professor of Economics and Psychology in the Social and Decision Sciences Department at Carnegie Mellon University He received his B.A. in economics magna cum laude from Brandeis University in 1977 and Ph.D. in economics from Yale University in 1985 with thesis titled Expectations and Intertemporal Choice.
The use of this is reference clearly captures and is consistent with the research. Gill is citing Loewensteins research on visceral filters. Gill is accurately captures/uses the research in his book.
Salvia, J., Algozzine, R., & Scheare, J.B. (1977). Attractiveness and school achievement. Journal of School Psychology, 15(1), 60-67.
30 others cited John Salvia is Associate Professor Department of Special Education Pennsylvania State University Robert Algozzine Assistant Professor of Special Education, University of Florida.
Gill accurately captures what the cited research stated (i.e. that teachers tend to favor attractive students)
Ambady, N., & Rosenthal, R. (1993). Half a minute: Predicting teacher evaluations from thin slices of nonverbal behavior and physical attractiveness. Journal of Personality and Social Psychology, 64(3), 431-441.
308 citations Article partly based on Nalini Ambady's doctoral dissertation conducted at Harvard University. Nalini Ambady was a social psychologist and a leading expert on nonverbal behavior and interpersonal perception. She was a Professor at Stanford University in the Department of Psychology. Rosenthal - Distinguished Professor Edgar Pierce Professor of Psychology, Emeritus Harvard University B.A. Psychology 1953 University of California, Los Angeles Ph.D. Psychology 1956 University of California, Los Angeles
The use of this citation was justifiable and appropriate.
4
Lord, K. R., & Burnkrant, R. E. (1993). Attention versus distraction: The interactive effect of program involvement and attentional devices on commercial programming. Journal of Advertising, 22(1), 47-60.
47 citations Kenneth R. Lord (Ph.D., The Ohio State University) is assistant professor of marketing, State University of New York at Buffalo. Robert E. Burnkrant (Ph.D., The University of Illinois) is professor of marketing at The Ohio State University.
Not used by Gill in the text of the chapter.
Ordonez, L. D., Schweitzer, M. E., Galinsky, A. D. & Bazerman, M. H. (2009b). On good scholarship, goal setting, and scholars gone wild. Academy of Management Perspectives, 23(3), 82-87
In reference section, but not in the text. I did not assess the credibility of this research.
Integrative Complexity: An Approach to Individuals and Groups as Information-Processing Systems Author(s): Michael J. Driver and Siegfried Streufert Source: Administrative Science Quarterly, Vol. 14, No. 2, Laboratory Studies of Experimental Organizations (Jun., 1969), pp. 272-285 Published by: Sage Publications, Inc. on behalf of the Johnson Graduate School of Management, Cornell University Stable URL: http://www.jstor.org/stable/2391105 Accessed: 25-03-2018 12:59 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Sage Publications, Inc., Johnson Graduate School of Management, Cornell University are collaborating with JSTOR to digitize, preserve and extend access to Administrative Science Quarterly
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
Michael J. Driver and Sieyfried Streufert
Integrative Complexity: An Approach to Individuals and Groups as Information-processing Systems
Individuals and groups can be viewed as information-processing systems which respond in a curvilinear fashion to three components of input load: complexity of information, noxity (unpleasantness) and eucity (pleasantness). An optimal input load is postulated, at which each system is expected to achieve maximum complexity in information-processing. At similar input levels, some systems are expected to show more complex information-processing than other systems.
Research is reviewed which suggests that the model holds for perception, information search, decision-making, and innovation. When productivity criteria are associated with complex information-processing, the model predicts productivity. A more complex phasic theory is then advanced, which argues that perceptual and decision-malking functions are separate and not synchronous.
One of the primary objectives of management science is to understand and control the group processes in highly productive organizations. In
this discussion productivity is viewed as strongly dependent on the information-processing char- acteristics of individuals, groups, and organiza- tions. Information can be regarded as the thread out of which all production decisions are woven. The particular way in which organi- zations search for information and handle it is therefore of great importance in developing a model for productivity.
This discussion outlines a new model of information-processing in systems of all types- individuals, groups, and organizations.1 Some recent experiments testing this model are pre-
sented and discussed, and a further extension of the model is then proposed.
THE INFORMATION-PROCESSING APPROACH TO PRODUCTIVITY
Definitions
First, it is useful to define key concepts. Of central concern is information. In Shannon & Weaver's (1949) information theory, this term has come to mean something which reduces uncertainty or entropy in a receiver. Informa-
tion has been quantified as "bits," where each
1 This paper is based on an invited address presented at the Institute of Management Science meeting, Feb- ruary 1966, Dallas, Texas.
bit means the reduction of uncertainty by one half. Although this technical view has yielded
some exciting results (Garner, 1962), it seems too restrictive and has led some (Schroder, Driver, and Streufert, 1967) to redefine infor- mation as anything that alters subjective (or objective) probabilities or utilities. Thus in this discussion, any environmental change which could cause a shift in expectations or evalua- tions will be termed information, regardless of whether certainty is decreased, increased or is unaffected, so long as utility (i.e., value times probability) is altered.
A second and crucial concept is system. The use of the term is based on von Bertalanffy (1952) and general systems theory (Miller, 1965). Von Bertalanffy, in his important treatise on biology, defined a system as a "complex of elements in mutual interaction." He proposed that all sets of organized elements follow certain general rules. These rules he termed general systems theory. In this dis- cussion, individuals, groups, and organizations are regarded as systems which possess certain general characteristics in common. It is recog- nized that there are serious dangers in making unwarranted analogies between systems of different levels of analysis, for biological, per- sonality, and social systems differ in many significant ways. Nevertheless, for present purposes, individual or personality and social systems are treated as having certain basic
272
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
Driver and Streufert: INFORMATION-PROCESSING SYSTEMS 273
similarities in the organization and structure of their parts, and are therefore expected to re- spond in basically similar fashion to certain aspects of the environment.
A third basic concept is input. Most gener- ally, input is any condition initially external to a system which upon contact with a system produces observable change in the organization or behavior of a system. Individuals, groups, and organizations are considered systems which operate to process inputs into outputs. The inputs which are of concern here are not defined
as energy units as in physical systems, but rather are defined as information units. An
information unit is any discrete condition of
the non-system environment capable of altering the utility or probability patterns in an indi- vidual or group. The nature of this information unit in any setting is arbitrary. One could define units as messages, sentences, words, or the presence or absence of a colored light de- pending on the refinement of measures of
change in the utilities or probabilities of the information-processing systems involved. In- dividuals, groups and organizations are con- ceived as information-processing systems, which respond in characteristic ways to variations in information inputs as they produce outputs.
The last concept is concerned with a crucial
aspect of input, the sheer complexity of the information. This is of central importance in
understanding systematic responses to varia- tions in input. Complexity of input can be defined as the number of utility or probability changes that an input can potentially evoke in a maximally open system as a result of the informational content of the input. A "maxi- mally open" system is one in which there are no mechanisms blocking the perception of any implications in the input. As will be seen, many information-processing systems "close out" in- formation and do not cope with the full com- plexity of the input.
Linear and Curvilinear Relationships Between Input and Output
There have been a number of attempts to develop models relating the complexity of the input to the functioning of human information- processing systems. One of the simplest and
earliest formulations was proposed by Herbert Spencer. Spencer (1897) suggested a simple linear function for relating the complexity of the
input to the complexity or organization of biological and social systems. He argued that the more complex the input, the more complex the system. In information-processing terms, this means that the complexity of the input directly determines the organizational capacity of a system to process information and produce complex outputs.
Some of the earlier research of Miller (1960) on information-processing in individuals and groups seems to offer partial support for Spen-
cer's view. Miller found that as he increased the complexity (i.e., number of units per time period and amount of information per unit) of information input, both individuals and groups
increased information-handling capacity linearly
up to an upper limit or asymptotic level. At the
asymptotic level, however, the systems began to omit inputs or let them accumulate in a
delayed action "list"-a process termed "queu- ing." It seemed, therefore, that each system had an upper limit for input beyond which Spencer's linear formulation was not valid. Some interesting work of Chapman, Kennedy, Newell, and Biel (1959) on a simulated air-defense system showed a similar, general rise in informa- tion-processing capacity to an asymptotic level.
More recently, Miller (1962) found that if he continued to raise the complexity of the input beyond the previously obtained asymp- totic level, the utilization of information began to decline in both quality and quantity. Chap- man et al. (1959) found a similar result in perception of input, if not in final output. This led Miller to consider some data which ranges from physiological systems to social organiza-
tion systems. For instance, Granit and Philips (1956) found that in the Purkinje cell of the retina, the response rate first increases with input, then decreases. Jung and Baumgartner (1955) found a similar curvilinear relation be- tween input complexity and the output of the optical cortex in the brain. At the individual person level, Quastler and Wulff (1955) found that if input complexity (i.e. number of notes per bar) was steadily increased, the quality of musical output in accomplished pianists first improved, but then deteriorated. Anderson and Fitts (1958) reported that three input dimen- sions were better than two or four in inducing maximum information transmission by human beings. At the group level, Lanzetta and Roby (1956) found an analogous curvilinear pattern
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
2 74 ADMINISTRATIVE SCIENCE QUARTERLY
with an optimal input level for group perform- ance. Finally, Fritz and Grier (1954) identified a similar curvilinear function between input
complexity and output quality in systems of
air-traffic control.
Many studies, however, do not show the
curvilinear relationship between input complex-
ity and system output. There are many possible
reasons for this deviation, but two factors have
been included in a new information-processing
model proposed by Schroder et al. (1967): (1) the nature of the system output and (2) the differences between systems in information- processing characteristics.
Relation of Integrative Complexity
to System Output
Schroder et al., suggested that not all output is systematically related to the complexity of the input. They proposed that it is only the integrative complexity in output that is related
to input complexity; their view is that informa- tion-processing systems can be described in terms of (1) the number of parts at work in a
system (i.e., roles in a group, ideas in a per- sonality, etc.), termed differentiation; and (2) the amount of interconnection between parts,
termed integrative complexity. The more a system is differentiated and integrated, that is, the more system parts are interrelated, the more
the potential information in the input is utilized in outputs. Output integrative complexity (e.g.,
the number of tactics relating to a goal) thus mirrors the integrative complexity of the parts in a system. As input complexity affects the
system's integrative complexity in a curvilinear
way it therefore relates to the integrative com- plexity of the system's output in a curvilinear
way. A high degree of integrative complexity in a system implies that outputs such as de- cisions, verbalized perceptions, or performance will reflect extensive prior use of information and that therefore these outputs will be char- acterized by a high degree of relatedness among parts. Integrative complexity of output is often what is meant by the quality of the output. Schroder et al. maintained that only when the quality of the output-or the degree of integra- tive complexity in the output-is assessed, can one obtain a clear curvilinear relationship with input complexity. Thus, for example, studies relating some simple response (e.g., conditioned eye blink), which requires little information
integration, to input complexity can hardly re-
sult in a curvilinear relationship. By contrast,
the studies to be described employ output mea-
sures that amply reflect internal changes in the
integrative complexity of the system caused by
change in input complexity. The integrative complexity model hypothe-
sizes that individuals and groups can be re-
garded as information-processing systems; that
the internal integrative complexity of a system varies in a curvilinear pattern as the complexity
of the input changes; and that therefore sys-
tems have optimal-input conditions (neither too
complex nor too simple), which maximize their
capacity for integrated complexity in output.
They further postulate that complexity of input
has three components: (1) the amount of
probability change implied (e.g., rate of infor-
mation input), referred to as pure information or load; (2) the amount of attainment of posi-
tive values (e.g., reward, referred to as eucity; and (3) the amount of attainment of negative values (e.g., threats), referred to as noxity. They argue that each of these three components
contributing to the complexity of the input bears a curvilinear relationship to the integra- tive complexity of a system and that they probably jointly affect information processing in a roughly additive fashion. A crucial feature
of this view is that the components, pure infor- mation load, eucity and noxity have inter-
changeable effects on information-processing
systems; e.g., threat and rate of information in the input could have similar effects on the
integrative complexity of systems. Walker
(1965) arrived at a very similar view of the relations among these variables after analyzing
a body of data collected by himself and others
on behavior of sub-human mammals. The basic integrative-complexity model is
summarized in Figure 1. The environmental input (input complexity) is a roughly additive function of information load, eucity, and noxity factors. At present, the evidence (e.g., Schroder et al., 1967) suggests that an additive relation- ship is found for at least part of the ranges of these variables, but further work will be needed to determine more precisely the interaction of these variables over all input ranges.
It will be seen in Figure 1 that the complexity of input into a system has an effect on system complexity in addition to its varied impact on system content (e.g., concept learning). If the
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
Driver and Streufert: INFORMATION-PROCESSING SYSTEMS 275
3.~ ~~~~~~~I
. ~ ~~~ . . ..
A _~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
IGR1FLWIGAOINEAIECPEIM6.
FIGURE 1. FLOW DIAGRAM OF INTEGRATIVE COMPLEXITY MODEL.
input complexity is moving away from the
optimal range of the system, the complexity of the system temporarily decreases. If the input is moving toward the optimal level, the com-
plexity of the system temporarily increases. If
the input is within the optimal range, the sys-
tem maintains operation at its maximum
capacity for complexity. Note that both the differentiation of parts and the integrative
complexity of the structure of the system are involved in system complexity.
Only two final points need be made in ref- erence to Figure 1. First, not all observable
behavior is a direct function of the complexity of the system. Much of the behavior of an individual or group system is most clearly related
to the content of the system, i.e., to past per-
sonal habits or group norms. However, when observable behavior strongly reflects integration of new information, e.g., in a decision for which
there are no precedents, then the complexity of a system is a significant factor in predicting
behavior. Furthermore, the optimal input complexity
for any system is not the same. Schroder et al. (1967) suggests that the past experience of an
individual or group system determines its unique
optimal level; and that current non-optimal input induces only temporary deviations in the complexity of the system away from the unique maximum associated with the optimal input
level of the system. However, it is suggested
that sustained deviations from the optimal input
complexity may cause permanent changes in the entire response range of a system, especially in regard to the maximum complexity which
the system can attain (see Schroder et al., 1967).
Differences in System Characteristics
The idea of different optimal input levels for individual or group systems leads to the second reason why a curvilinear relationship is not always observed between system input and out- put. Schroder et al. (1967) argue that systems have different optimum input ranges which re- flect different capacities for maximum complex- ity. Furthermore, as input moves away from
the optimum, each system responds with a characteristic temporary loss in integrative complexity. Each system then has its own curvilinear pattern of change in integrative complexity, which is consistent and reliable over time. The curvilinear pattern of reaction
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
276 ADMINISTRATIVE SCIENCE QUARTERLY
to complexity in input is called the character-
istic level of integrative complexity in the
system. This idea is quite distinct from older
"trait" theories of social science in that an individual, group, or organization is not ex-
pected to display the same amount of system
complexity in overt behavior in all situations; rather, each system should react with a predict-
able curvilinear pattern of system complexity as the complexity of input changes. In general,
a system with a high characteristic level of
complexity shall be referred as "complex" and a system with a low characteristic level of complexity shall be referred to as "simple."
Information-processing systems can thus be
classified as displaying a characteristic cur- vilinear pattern of integrative complexity across the entire spectrum of complexity in input con-
ditions. There is therefore not one curvilinear
pattern relating input and output, but rather a
whole family of such curves. Figure 2 repre-
sents two of these curves, showing two systems
HIGH
Z OPTIMAL INPUT 0 LL]SYSTEM A
(LOW\ ( HIGH CHARACTERISTIC Z / \ ~~~~~~~~~~~~~~LEVEL OF INTEGRATIVE
< COMPCIY TY)
TIVE . (LOW CHARACTERISTIC of different characteristicLEVEL OF INTEGRATE ite
cc~~~EVROMNA NU COMPLEXITY)
complexity. Systemr A is a more complex system than System B. Systems A and B could be
individual or group systems; the same dynamics
should hold in either case. The curves in Figure
2 are based on actual research findings (Streu-
fert and Schroder, 1965; Streufert and Driver, 1965; Driver, 1962; 1969).
The data suggests that all information-
processing systems show roughly symmetrical
curvilinear reactions to input, and that more complex systems (e.g., A in Figure 2) are al- ways more complex at comparable inputs than simple systems (e.g., B), except at extreme inputs. The data also suggests that at very low
or high inputs, systems operate at extremely simple, often one-dimensional levels (Driver, 1962; 1965; 1969). The curves of Figure 2 are found in the research literature if one varies either load or noxity-eucity, and holds the other
variable constant. Their additive effect is only an operating assumption, which may yield to a more complex interaction formulation with new studies.
Figure 2 shows a possible source of confusion in information-processing literature. If systems A and B are compared at different input condi- tions (e.g., A at point 1 and B at point 2 in Figure 2) B may appear to be more complex in output than A. Some of the confusion in the literature may stem from the mixing of systems with very different curve patterns or from a lack of comparability of input conditions.
It must be emphasized that the integrative
complexity model does not equate individuals and groups on all aspects. It only states that
both types of system can be characterized as showing more or less integrative complexity, both will respond in a curvilinear fashion to input, and both have characteristic levels of integrative complexity that represent their unique curvilinear response to input. Further- more, individual and group levels of systems interact. Integrative complexity in individuals
tends to produce integrative complexity in groups (Tuckman, 1964; Stager, 1967), but
the relationship between group complexity and individuals is not simple. It is probable that if
group structure is slightly more complex than
the personality structures of the individuals
comprising the group, some increase in the com-
plexity of the personality structures may occur.
Also if the group structure is simpler or far
more complex than its members' individual
structures the complexity in individual per-
sonality structure may decrease. Far more re-
search is needed before the relationship between
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
Driver and Streufert: INFORMATION-PROCESSING SYSTEMS 277
the characteristics of group and individual systems can be fully resolved.
Motivation
Besides the capacity to handle input, infor-
mation-processing systems are also characterized
by their motivation to seek and use information.
Recent trends in motivational analysis (White, 1958; Hunt, 1963) suggest that classic motiva-
tional constructs such as drives (e.g., hunger) and affects (e.g., fear) no longer can be re-
garded as primary sources of all behavior. For
instance, one animal study (Montgomery, 1952)
found that hunger arousal actually diminished exploratory, information-seeking behavior, while another study (Montgomery and Monk- man, 1955) revealed that fear had similar
inhibitory effects. Clearly the arousal of strong
drives or emotions is not motivating this infor- mation oriented behavior as classic motivation
theory would suggest. Yet, animals and human beings clearly work and learn merely to obtain information. In one classic study, Butler (1957) found that monkeys would learn and persist at the most difficult problems, merely to be afforded the dubious pleasure of looking at the experimental laboratory through a peephole!
This kind of data has led Hunt (1963) and others to postulate biological "cognitive moti- vations" inherent in information-processing itself. Cognitive motivation is not derived from
other motivations such as drives. The motiva- tion to seek a certain optimal amount of com- plexity in one's environment is suggested as one
such cognitive motivation. This motivation has a sound basis in animal data (Denenberg, 1964;
Walker, 1965) and can be suggested as a con- comitant of various levels of integrative com- plexity as defined in the integrative-complexity
model. It can be argued that each information- processing system has an inherent motivation
to maintain input conditions at an optimal level, which maximizes its internal integrative com- plexity. Thus, the model suggests that systems differ not only in information-processing capaci- ties but also in motivation for a particular amount of input complexity. Although there
are supporting data for individual human and
infrahuman systems (Bryson and Driver, 1969), there are few data on the preference for com- plexity by groups. Here again one must use
caution in making analogies from individuals to groups.
RELATION OF PRODUCTIVITY TO COMPLEXITY OF OUTPUT
To link productivity to the integrative com-
plexity model, those system outputs which reflect variable integrative complexity in sys-
tems and contribute to various conceptions of
productivity must be considered. We will
examine such outputs as: the amount of in-
formation used in perceptions, the amount of complexity in decisions, the amount of com- munication, the extent of information search, and the degree of innovation in behavior.
Information Processing and
Group Productivity
Perception. Two studies were particularly
concerned with perception (Streufert and Driver, 1965; Driver, 1962; 1969). One deals
with the effects of pure information load on perceptual complexity, while the other con- siders the effects of noxity-eucity variables on perception. The basic finding was that the
pattern illustrated in Figure 2 was substantiated
whether load or noxity-eucity was varied. In both studies perceptual complexity varied in a curvilinear fashion with increasing input load.
In Streufert and Driver's study, information load was varied by controlling the number of messages per half hour in a tactical game de- veloped by Streufert et al. (1965a). The game confronts two teams in a military negotiation situation analogous to Viet Nam. Perceptions of the other team were filled out by each team and rated for complexity using methods de- scribed by Schroder et al. (1967). Some factors considered were: empathy, cause-effect pat-
terns, three (or more)-step inferences, and imputation of intentions to others. While noxity-eucity factors (e.g., success-failure) were held constant, perceptual complexity varied in
a curvilinear, symmetrical way with input load.
The complexity (i.e., characteristic level of
complexity) of the subjects in the tactical game
was assessed using projective measures described by Schroder et al. (1967). For example, in the method of Impression Formation, subjects wrote
a one page description of a "person" who was
described to them only by 3 pairs of potentially
conflicting adjectives (e.g., intelligent-stubborn,
industrious-envious, etc.). The descriptions
could be rated for general level of integrative
complexity using a manual developed by Schro-
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
278 ADMINISTRATIVE SCIENCE QUARTERLY
der et al. (1967) and by Streufert and Driver
(1967). In the tactical game, half of the subjects
were complex (i.e., high characteristic levels
of integrative complexity) and half were simple
as measured by scores on tests such as the Im-
pression Formation Task. Intelligence test
scores showed no significant differences between
these two groups in I.Q. Both groups showed
curvilinear reactions to! load; however, at all
comparable load conditions, the high-complexity
group displayed more integration in perception
than the low-complexity group. The empirical
curves resembled those of Figure 2. Both groups
peaked at the same optimal input level. The differences between groups were greatest at the optimal input and decreased as lower or higher loads were approached. More recent work in
progress suggests that the simultaneous peaks
of high and low-complexity systems in data
from the tactical game are due to unrefined
measurement. More refined measurement cur-
rently suggests that simpler systems may peak
at less complex input conditions than complex systems.
The second perception study (Driver, 1962; 1965; 1969) found basically similar results
when noxity-eucity was varied, with input load kept relatively constant. In this study seven teams of players simultaneously operated
simulated "nations" in Guetzkow's Internation
Simulation (Guetzkow et al., 1963). A run consisted of seven teams operating for 5 days; there were 17 separate runs of the simulation. In two runs, the communications data indicated low eucity (interest). The remaining 15 runs were scaled for threat (noxity) level using an analysis of armament and treaty data. Thus the runs formed a continuum from low interest (eucity), to high interest and low threat, to high interest and high threat.
For each run the average perceptions of the 7 simulated nations by the players were assessed using multidimensional scaling (Jackson and Messick, 1963), a mathematical technique for uncovering the basic perceptual dimensions integrated in judgments of general similarity among stimuli. The average perceptions so scaled, showed both the differentiation (number of dimensions) and integration (weighting pat- tern of dimensions) of perception in each run.
Using either differentiation or integration as a measure of complexity, a curvilinear pattern was
found between the average perceptual com-
plexity and general interest-threat level. Maxi- mum complexity was found in runs with high interest and relatively low threat. Threat was,
in general, negatively related to perceptual com-
plexity. Subjects were divided into groups of
high and low characteristic levels of integrative
complexity using methods of selection similar to those described for the tactical game. Again it was found using multidimensional scaling
that the perceptions of the more integratively
complex subjects were more complex at all levels of input noxity-eucity except the most
extreme.
These two studies illustrate the basic integra- tive complexity model quite well. Not only do they show how perception-the gateway to all information processing-is affected by input
complexity and the integrative complexity of group members, but they also reveal the ap- parent interchangeability of the components contributing to input complexity (e.g., threat- interest and input load).
Information search. Besides perception, in-
formation search behavior can also be related to the model. Information search may be simple, defined as a request by the subject to some external agency for more information; or complex, defined as the use by the subject of some strategy upon the environment to generate information.
The research showed a high level of both kinds of search with low input loads. Increased input loads related to decreased search. The
relationship between search and noxity-eucity is not yet known. The linear relation in search
behavior demonstrates that it is only partially controlled by the complexity of a system. Particularly at low input levels, content factors, such as social desirability, operate to sustain high search levels. It is only at higher loads that the resultant simplification of individual or group information processing structures over- rides content factors and forces the abandon-
ment of search as one means of reducing the overload on a system.
A difference in search behavior between complex and simple individual systems is also found. In general both types of search are em- ployed by high-complexity and low-complexity systems at low input loads, but the complex system tends to use complex search tactics, while the simpler system tends to use the simple
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
Driver and Streufert: INFORMATION-PROCESSING SYSTEMS 279
search tactic. This bias becomes very marked
under high load conditions, when the nonfavored
tactic drops out of use much more readily than
the favored tactic.
The most complete data on search behavior
are found in a study (Streufert et al., 1965b)
using the tactical game described by Streufert
et al. (1965a). Two search tactics were per-
mitted: simple search-requests for informa-
tion from "intelligence" sources, and complex
search-orders by players to initiate reconnais-
sance-type missions by their own units that
might provoke useful responses from the "enemy." Input load was varied by manipulat- ing message rate. Half the players were of
complex characteristic levels, half were simple. Measurement of individual complexity was
handled as described above.
The results revealed a general decrease in
search from very high to moderately high levels as input load increased. At no time was search abandoned. Complex players maintained rela-
tively high levels of complex search, particularly at high input load; they rapidly abandoned simpler search at high input loads, although they made high use of simple search at moderate loads. Simpler players favored simple search,
particularly at higher input loads. Each type of system seems to maintain a search tactic
suited to its complexity level as long as possible, while using the other tactic as a supplement. Note, however, that simple systems used the supplementary complex strategy most at lowest input loads, while complex systems used their
supplementary strategy most at moderate input -possibly indicating that maximum search is attained by complex and simple systems at dif- ferent input loads.
It appears that preferred complexity of search
tactics varies directly with the complexity of systems. In a study of simulated stock market
behavior by Driver (1960), teams composed of subjects of high and low complexity operated
brokerage firms in a simulated market. Several sources of information were available, includ- ing an investment analyst's report that provided
a simple, quasi-authoritarian method of obtain- ing output instead of more complex self-initiated
methods. Subjects were asked to rate their preference for information sources. The less
complex subjects significantly preferred the simpler input source (analyst report), confirm- ing the hypothesis that different complexity
levels display preferred modes of information
search.
Communication and innovative interaction.
Complex, multilateral communication efforts in
a social situation may be regarded as a form of
complex information search. In the Internation
Simulation study, cited above, it was found that communication varied as a function of the char-
acteristic level of integrative complexity of
decision makers in the groups (Schroder et al., 1967). Communication was assessed by count-
ing the number of press releases given to the
newspaper published in each simulation run.
There was a significant tendency for the more
complex systems to send out more press releases,
regardless of differences in input conditions.
The suggestion is that communication may be a
still more complex search tactic than the self-
initiated tactics of the tactical negotiation game,
and therefore emerges even more clearly as a
tactic of complex systems.
Another tactic analyzed in the Internation
Simulation is the initiation of trade relations with simulated nations outside the nation alli- ance "blocs," which were defined at the game's beginning by the experimenter. It was found that, except for groups which had high levels of military-economic power, the groups com- posed of more complex subjects engaged in significantly more trade outside defined blocs than groups with less complex members. This result suggests that another complex communi- cation and information-gaining tactic-setting up general trade relations-is a more typical tactic of complex systems. This result also suggests that more complex systems are more likely to engage in innovative, creative tactics in dealing with input. So long as one considers tactics dictated or sanctioned by an external source or authority, one finds either no dif- ference between complex and simple systems in behavior (as for example in trade within the
pre-established blocks) or a tendency for simpler systems to produce more of the output sanctioned by authority (e.g., delegated search). However, when tactics require original, complex, self-initiated, innovative behavior which may ignore or violate authority, complex systems
seem more adapted to the activity and generate
more of this kind of output. For example, Tuck- man (1966) recently found a strong positive
correlation between innovative behavior and
characteristic levels of complexity.
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
280 ADMINISTRATIVE SCIENCE QUARTERLY
Decision-making. Once information has been found and perceived, how is it used in decisions? In the tactical game, Streufert and Schroder (1965) and Streufert, Suedfeld, and Driver (1965b) studied the use of information in de- cisions by assessing the number of input units that subjects integrate in a decision. For every input (e.g., message), subjects indicated the decisions to which it was relevant. The basic curvilinear pattern was again found in both of these studies. A rising input load first increased, then decreased integration of information in decision-making. At all but extreme input con- ditions, the more complex systems integrated far more input than simpler systems. These data, along with the perceptual data already dis- cussed, clearly indicate that although simpler systems may at times engage in more input search (particularly of a simple, authority oriented variety) they usually do not integrate more information into their output.
Prolductivity. It might seem that since more complex systems engage in more complex tactics of information search, perceive more input, and use more information in decisions, they would generally be more productive than simpler systems. This is not so. The critical issue here is the criterion used to measure productivity. If the criterion reflects behavior that demands
complex search, high levels of perception of input, and complex integrations in decisions, then clearly more complex systems are at an advantage. If on the other hand, the criterion is such that simpler search methods and the use of limited information is required for achiev- ing the goal, then there may be no; difference or even an advantage for simpler systems.
Some data from the stock market simulation and Internation Simulation support these views. In one version of the stock market game (Driver, 1960), there were two performance criteria: one simply required adherence to simple information search, while the other re- quired use of more complex information and self-initiated activity. Although the results were not always quite significant, the trends were clear: more complex systems did better on the criterion requiring complex behavior, while less complex systems did better on the criterion requiring simpler behavior.
These findings were amplified by two studies in which simple and complex criteria were used separately. Tuckman (1964) used a different
version of the stock market game in which the criterion for productivity reflected only simple information tracking and found that subjects with low integrative complexity had an advan- tage. By contrast, in the Internation Simula- tion, where the criterion for performance generally requires subtle and complex informa- tion processing, comparisons of high and low complexity systems, which were in blocs com- posed predominantly of their own complexity level, showed that the complex systems were more productive. Bloc membership was found to affect productivity strongly. Complex groups in blocs composed of predominantly simple groups showed depressed productivity, while
simple groups in generally complex blocs showed elevated performance. Average pro- ductivity in the Internation Simulation had a curvilinear relationship to input load, in keeping with the integrative complexity model. In sum- mary, it seems that only when the criterion reflects integrative complexity in output will the model presented here be relevant.
Implications of Model for Management
Three implications seem important for man- agement:
1. The tasks involved in job assignments of individuals and groups should be matched to
the characteristic level of integrative complexity of the incumbents. It would be unproductive to place a very low complexity person (or group) in a task for which complex information processing is the criterion for success. Such an incumbent might collapse under the pressure, try to reduce the load (via staffs), or redefine the position which could have injurious reper- cussions throughout the organization. Equally unproductive would be an assignment of a high complexity person (or group) to a task which required simple, routine behavior for success. The results might be either total withdrawal to internal processes (day-dreaming) and a consequent decrease in performance or efforts to complicate or involve the assignment, which could impair performance and lead to other organization strains.
Matching of jobs and personnel complexity clearly requires measurement of both. Some preliminary measures of job demands are being developed and several measures of characteristic levels of integrative complexity are available for individuals. Schroder et al. report two
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
Driver and Streufert: INFORMATION-PROCESSING SYSTEMS 281
projective measures of personality complexity- the Impression Formation Test described above and the Paragraph Completion Measure. The latter consists of a series of words such as parents, roles, and others which the subject uses as a topic for a short paragraph expressing his reaction to the topic. Both projective measures have scoring manuals to establish a unified scoring method, and with proper training, a rater can attain high reliability (.8-.95 inter- rater reliability).
A more objectively scored test of complexity is desirable, however, and Schroder et al. de- scribe several psychometric techniques which could be used to assess complexity in specific content areas. The authors are currently at work on two general objective measures of characteristic levels of integrative complexity. One is nearly complete and is being evaluated in a variety of experimental, educational, and industrial settings. This test-the Purdue-
Rutgers Experience Inventory I-has extremely high internal consistency (e.g., its factor struc- ture in two separate analyses has reflected the theoretically expected relationships among items very well) and has good item reliability (.4- .75). Its validity as a predictive instrument is being ascertained in current research. Prelim- inary results of one analysis show this test is related strongly to established measures of teacher classroom style. This test is applicable to business management settings but the second test, the Integration Style Test, which deals with business problems should prove even more useful in such settings. The Integration Style Test has proven reliable across several content areas of business decisions and also reveals via factor analysis a theoretically meaningful in- ternal structure.
2. A second implication is that for any group or individual, one should maintain the optimum input range if maximum output is desired. This implies the use of even nonfunctional informa- tion inputs or emotional inputs (e.g., slight threats or moderate rewards) to maintain maxi- mal levels in systems if input is too low; also, it implies the need for periodic relief from over- load positions (possibly by job rotation, vaca- tions, etc.). The problem is that if systems are allowed to decline because of low input level or if they are constantly overloaded (e.g., certain executives) they are in danger of a permanent regression to less and less complex characteristic
levels of integrative complexity. This would
mean that the capacity of organizations to meet problems requiring integratively complex sys-
tems is decreased. Often individual or group
systems which have regressed are simply re- placed. But this seems less efficient and more costly than rotating assignments or varying load in such a way that system complexity is not decreased. Whether the optimal input level is the same across systems is an unresolved
problem.
3. A final intriguing problem concerns meth- ods of enhancing integrative levels, if desired.
Dember's (1960) concept that a "pacer" stimu- lus, just a little ahead of a system's optimal input complexity, produces an increase in sys- tem level and possibly optimal motivation is
well worth testing in human systems. Careful planning might enable an organization to pro- vide just that amount of input above a person's optimum to obtain a gradual increase in integra- tive complexity levels. It should be clear, however, that for some criteria, increasing com-
plexity of systems is not desirable.
A PHASIC THEORY
The model presented was oversimplified in order to obtain some rough preliminary chart- ings in what is largely an unexplored area. It was conveniently assumed that all parts of an information-processing system function in syn- chrony and that all parts follow a basically curvilinear pattern; however, it is already clear that some forms of information search do not fit such a pattern. A closer look at the data suggests a phasic model of information process- ing.
A basic postulate in the phasic model is that there are at least two subsystems in an informa- tion-processing system. One subsystem is termed the perceptual subsystem and is largely concerned with data search and intake. Intake in a system is that set of processes which code, interrelate, evaluate (especially in relation to need for immediate action) and store input. This subsystem transmits processed input to the second subsystem-the executive subsystem, which translates input into action decisions,
actions, and strategies. For the individual system, this concept has already been developed by Pribram (1958) in his analysis of brain functioning. Pribram suggests that the brain
operates in terms of two anatomically distinct
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
282 ADMINISTRATIVE SCIENCE QUARTERLY
HIGH
Perceptual Subsystem Complexity
_ *-s-s _ Perceptual Subsystem Total Output Quantity
? ~ ~ ~ ~ ~ ~ = Executive Subsystem Complexity o go 0 0 0 0 0 0 Executive Subsystem Total Output Quantity _ go
z ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~a0*
0~~~~~~~~~~~~~~~~~~~~~~~~~~~g
CL~~~~~~~~NIOMNA INPU COMPLEXIT
0
-j 4 <
o- 0
FIGURE 3. INTEGRATIVE COMPLEXITY OR OUTPUT QUANTITY OF PERCEPTURAL AND EXECUTIVE
SUBSYSTEMS IN RELATION TO ENVIRONMENTAL INPUT COMPLEXITY.
(research groups) or input intake (executive secretaries), whereas other segments are con-
cerned with executive output functions (sales or production managers).
The next basic postulate in the phasic model is that the two (or more) subsystems of an information-processing system are not syn-
chronous in their response to input. Both sub- systems may respond to increasing load in a series of interdependent phases, as shown in Figure 3. Phase 1 represents the period in
which both perceptual and executive subsystems increase in integrative complexity as conditions move from relatively low input to moderate input. In this phase the perceptual system may
reach its normal peak level of complexity. Some data (Schroder et al., 1967) suggest that some- times the perceptual system peaks at lower input values than the executive system (which
peaks in phase 3), while at other times their peaks coincide. The difference in peaking may be due to the filtering action of the perceptual subsystem, which does not transmit all input
intrinsic information-processing systems. The
frontal system appears to process planning,
decisions, and action, while the posterior
system handles perceptions and judgments.
Judgment refers to the assignment of input
to evaluative categories (e.g., I like x);
perception refers to the assignment of input to
reality description or relational categories (e.g., x is red; x causes y). Conversely, decisions involve the integration of perceived inputs and
stored behaviors to attain some objectives
(based on judgments of the worth of various goals). Plans are sets of behaviors developed to implement decisions.
It is possible to extend Pribram's notions to
groups and organizations as well, always keep-
ing in mind that individual and group systems
are comparable only as information-processing
systems. In many other significant ways, bio- logical and social systems may be quite dif-
ferent. Nevertheless, as with individuals, one
can examine groups and often find segments of
the group system focusing on information search
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
Driver and Streufert: INFORMVIATION-PROCESSING SYSTEMS 283
received and so reduces input for the executive
subsystem.
Phase 2 represents data (Chapman et al., 1959) which suggests that at moderate inputs,
exceptionally complex integrations of perceptual information may occur, which raise the percep-
tual subsystem to supernormal levels. If one
assumes only a limited capacity in energy (re- sources) or processing units in a system, the cost of this supernormal perception may be a
drop in the energy or capacity of the executive system, resulting in a temporary drop in its complexity.
Phase 3 shows a decline in the integrative complexity of the perceptual subsystem, which begins to omit input altogether (Quastler and Wulff, 1955; Miller, 1960; 1962). However, if the rate of input omission in the perceptual
subsystem is assumed to be less than the rate of new imput, the executive subsystem still re- ceives more input than in phase 2 and is stimu- lated to attain its peak complexity. Suppose, for instance, input rate corresponding to phase 2 is 10 units per time unit. As this input rate increases to 20, it may force an overload decline in the perceptual subsystem from transmission
of 100% of input to 80%. Futhermore, the amount of stored information added to input and thus transmitted to the executive subsystem (i.e., integratively complex perception) will
also decline. Even so, the perceptual subsystem will transmit 80% of 20 inputs (equals 16 in- puts) to the executive subsystem, whereas at
lower input levels only 10 inputs were trans- mitted. Thus, despite lower capacity in the
perceptual subsystem, the executive subsystem may still receive a greater load than in phase 2 and so attain maximum complexity in this phase.
In phase 4, under still greater input, it ap-
pears that the executive subsystem may at first maintain its integrative complexity and put out
a constant number of integratively complex decisions, while meeting further input increases with a rise in simple outputs (Streufert, Driver, and Hahn, 1967). Thus, while the number of complex decisions stays constant, reflecting a steady state of subsystem complexity (dashed line in Figure 3), the number of simple decisions increases resulting in an increase in the total
number or quantity of outputs from the execu-
tive subsystem (dotted line in Figure 3). How- ever, in phase 5, the number of integratively
complex executive outputs begins to fall as the
complexity of the executive subsystem declines; while simple output continues to rise resulting in a continued rise in total output. Finally in
phase 6, even simple outputs begin to decline,
hence total output declines, as the information-
processing system reaches exhaustion-energy,
resources, and even processing units are worn
out at described in the classic work on stress by Selye (1956).
It should be noted that whereas perceptual
complexity decreases from phase 3 on, the
amount of input perceptually processed and
transmitted (with no complex additions from
memory) steadily increases, despite increasing
omissions. An alternate model might be that
in some systems the rate of omission of percep-
tual input or the loss of information due to the
failure of the perceptual subsystem to perform
complex judgments or perceptions exceeds the
rate of increase in input. This effect could
produce states of low input in the executive
subsystem, which would produce phase 1-3 type
action in the executive subsystem even under
very high inputs.
It should be evident that this is only a pre- liminary sketch of a potentially more complex information-processing model which could be
applied to both simple and complex systems. Clearly systematic research is needed to evalu- ate each phase of the proposed sequence. For instance, the phasic model might give one a
very precise indication of how much input will produce an above-optimal perceptual response
(i.e., phase 2) rather than an optimal executive response (i.e., phase 3). At an organizational level it may be found that some individuals excel in perceptual data handling functions, while others excel in the executive functions. These possibilities are being explored.
It is hoped that these models and studies can provide a useful perspective on the problems of
management science. It is recognized that many factors, some of which may seem crucial, have not been dealt with, but this is a preliminary
attempt at a synthesis of information-processing approaches to management.
Michael J. Driver is associate professor of management at the University of Southern Cali-
fornia. Siegfried Streufert is associate professor of psychology at Purdue University.
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
284 ADMINISTRATIVE SCIENCE QUARTERLY
REFERENCES
Anderson, M. S., and P. M. Fitts
1958 "Amount of information gained during
brief exposure of numerals and colors." Journal of Experimental Psychology, 56: 362-369.
Bryson, J., and M. Driver
1969 "Conceptual complexity and internal
arousal. Psychonomic Science, in press.
Butler, R. A.
1957 "Incentive conditions which influence
visual exploration." Journal of Experi- mental Psychology, 54: 345-352.
Chapman, R. C., J. L. Kennedy, A. Newell, and W. C. Biel
1959 "The system research laboratory's air de-
fense experiments." Management Science, 5: 250-269.
Denenberg, V.
1964 "Critical periods, stimulus input and emo-
tional reactivity." Psychological Review, 71: 335-351.
Dember, W. N.
1960 Psychology of Perception. New York: Holt-Dryden.
Driver, M. J. 1960 "The relationship between abstractness of
conceptual functioning and group per- formance in a complex decision making environment." Master's thesis, Princeton
University. Driver, M. J.
1962 "Conceptual structure and group processes in an internation simulation. Part one: the perception of simulated nations." Edu- cational Testing Service Research Bulle- tin: 62-15. Princeton, New Jersey.
Driver, M. J. 1965 A Structural Analysis of Aggression,
Stress and Personality in an Internation Simulation. Working paper, Purdue Uni- versity, Institute for Research in the Be- havioral, Economic and Management Sci- ences, Paper No. 97.
Driver, M. J. 1969 Threat, Communication Rate and the
Complexity of Social Perception. Paper given at Western Psychological Associa- tion meeting, Vancouver, British Columbia.
Fritz, E. L., and G. W. Grier 1954 Report No. R-54. Urbana: Control Sys-
tems Laboratory, University of Illinois. Garner, W.
1962 Uncertainty and Structure as Psychologi- cal Concepts. New York: Wiley.
Granit, R., and C. G. Philips 1956 "Excitatory and inhibitory processes act-
ing upon individual Purkinje cells of the
cerebellum in cats." Journal of Physiol- ogy, 133: 520.
Guetzkow, H., C. Alger, R. Brody, R. Noel, and R. Snyder
1963 Simulation in International Relations. Englewood Cliffs, New Jersey: Prentice-
Hall.
Hunt, J. McV.
1963 "Motivation inherent in information proc-
essing and action." In 0. J. Harvey (ed.),
Motivation and Social Interaction: 50-73. New York: Ronald.
Jackson, D., and S. Messick
1963 "Individual differences in social percep- tion." British Journal of Social and Clini-
cal Psychology, 2: 1-10.
Jung, R., and G. Baumgartner 1955 "Hemmungsmechanism und bremsende
stabilizierung an einzelnen Neuronen des
optischen Cortex." Pflugers Archiv fur die gesamte Physiologie des Menschen und der Tiere, 261: 434-456.
Lanzetta, J. 1963 "Information acquisition in decision mak-
ing." In 0. J. Harvey (ed.), Motivation and Social Interaction: 239-260. New York: Ronald.
Lanzetta, J., and T. Roby 1963 "Effects of work group structure and cer-
tain task variables on group performance." Journal of Abnormal and Social Psychol- ogy, 53: 307-314.
Milburn, T. W., and J. F. Milburn 1965 Predictions of Threats and Beliefs about
How to Meet Them. Working paper,
Northwestern University, Department of Psychology.
Miller, J. G. 1960 "Information input overload and psycho-
pathology." American Journal of Psychia- try, 116: 695-704.
Miller, J. G. 1962 "Information input overload." 6th An-
nual Report: 30-31. Ann Arbor: Mental Health Research Institute.
Miller, J. G. 1965 "Living systems." Behavioral Science, 10:
337-412.
Montgomery, K. 1952 "A test of two explanations of spontane-
ous alternation." Journal of Comparative and Physiological Psychology, 45: 287- 293.
Montgomery, K., and J. Monkman
1955 "The relation between fear and explora-
tory behavior." Journal of Comparative and Physiological Psychology, 48: 254- 260.
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
Driver and Streufert: INFORMATION-PROCESSING SYSTEMS 285
Pribram, K. H. 1958 "Neocortical functions in behavior." In
H. Harlow and C. Woolsey (eds.), Bio- logical and Biochemical Bases of Be- havior: 151-172. Madison: University of
Wisconsin Press.
Quastler, H., and J. I. Wulff 1955 Report R-62. Urbana: Control Systems
Laboratory, University of Illinois.
Schroder, H., M. Driver, and S. Streufert 1967 Human Information Processing. New
York: Holt, Rinehart and Winston. Selye, H.
1956 The Stress of Life. New York: McGraw- Hill.
Shannon, C., and W. Weaver 1949 The Mathematical Theory of Communica-
tion. Urbana: University of Illinois Press. Spencer, H.
1897 The Principles of Psychology. New York: Appleton-Century.
Stager, P. 1967 "Conceptual level as a composition vari-
able in small group decision making." Journal of Personality and Social Psy- chology, 5: 152-161.
Streufert, S., M. Clardy, M. Driver, M. Karlins, H. Schroder, and R. Suedfeld
1965a "A tactical game for the analysis of com- plex decision making in individuals and groups." Psychological Reports, 17: 723- 729.
Streufert, S., and M. Driver 1965 "Conceptual structure, information load
and perceptual complexity." Psychonomic Science, 3: 249-250.
Streufert, S., and M. Driver 1967 "Impression formation as a measure of
conceptual structure." Educational and Psychological Measurement, 27: 102 5- 1039.
Streufert, S., M. Driver, and K. Hahn 1967 "Components of response in complex de-
cision making." Journal of Experimental Social Psychology, 3: 286-295.
Streufert, S., and H. Schroder 1965 "Conceptual structure, environmental
complexity and task performance." Jour- nal of Experimental Research in Personal- ity. 1: 132-137.
Streufert, S., P. Suedfeld, and M. Driver 1965b "Conceptual structure, information search
and information utilization." Journal of Personality and Social Psychology, 2: 736-740.
Tuckman, B.
1964 "Personality structure, group composition and group functioning." Sociometry, 27: 469-487.
Tuckman, B.
1966 "Integrative complexity: its measurement and relation to creativity." Educational and Psychological Measurement, 26: 369- 382.
Von Bertalanffy, L. 1952 Problems of Life. New York: Harper.
Walker, E. L. 1964 "Psychological complexity as a basis for
a theory o.f motivation and choice." In D. Levine (ed.), Nebraska Symposium on Motivation- 1964: 47-99. Lincoln: Uni- versity of Nebraska Press.
White, R. W. 1958 "Motivation reconsidered: The concept
of competence." Psychological Review, 66: 297-333.
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 12:59:10 UTC All use subject to http://about.jstor.org/terms
ORGANIZATIONAL BEHAVIOR AND HUMAN DECISION PROCESSES
Vol. 65, No. 3, March, pp. 272 – 292, 1996 ARTICLE NO. 0028
Out of Control: Visceral Influences on Behavior
GEORGE LOEWENSTEIN
Carnegie Mellon University
I. INTRODUCTION Understanding discrepancies between behavior and
perceived self-interest has been one of the major, but Avrum Goldstein, in his instant classic Addiction, largely untackled, theoretical challenges confronting
provides the following account of relapse to drug addic- decision theory from its infancy to the present. People
tion:often act against their self-interest in full knowledge that they are doing so; they experience a feeling of being Relapse is, of course, always preceded by a decision to use, ‘‘out of control.’’ This paper attributes this phenomenon however vague and inchoate that decision may be. It is an to the operation of ‘‘visceral factors,’’ which include impulsive decision, not a rational one; and it is provoked by drive states such as hunger, thirst and sexual desire, craving — the intense and overwhelming desire to use the moods and emotions, physical pain, and craving for a drug (1994, p. 220). drug one is addicted to. The defining characteristics of
Goldstein is anxious to portray relapse as a decisionvisceral factors are, first, a direct hedonic impact (which is usually negative), and second, an effect on the relative involving personal volition, to bolster his position that desirability of different goods and actions. The largely drug users should be held personally accountable for aversive experience of hunger, for example, affects the their behavior. However, the difficulty of doing so is evi- desirability of eating, but also of other activities such as dent from his resorting to adjectives such as ‘‘impulsive’’ sex. Likewise, fear and pain are both aversive, and both and ‘‘inchoate’’ to describe the decision and his picture increase the desirability of withdrawal behaviors. The
of craving as ‘‘intense’’ and ‘‘overwhelming.’’ The addictvisceral factor perspective has two central premises: knows, in one sense, that taking the drug is the wrongFirst, immediately experienced visceral factors have a course of action but is unable to translate this belief intodisproportionate effect on behavior and tend to ‘‘crowd action. Craving, it seems, has the capacity to drive aout’’ virtually all goals other than that of mitigating the
visceral factor. Second, people underweigh, or even ig- wedge between perceived self-interest and behavior. nore, visceral factors that they will experience in the Understanding discrepancies between self-interest future, have experienced in the past, or that are experi- and behavior has been one of the major, but largely enced by other people. The paper details these two as- untackled, theoretical challenges confronting deci- sumptions, then shows how they can help to explain a sion theory from its infancy to the present (though, wide range of phenomena: impulsivity and self-control,
see Beach, 1990; Fishbein & Azjen, 1975; Janis &drug addiction, various anomalies concerning sexual be- Mann, 1977; Kuhl & Beckmann, 1994). In 1960,havior, the effect of vividness on decision making, and Miller, Galanter, and Pribram lamented that ‘‘some-certain phenomena relating to motivation and action. thing is needed to bridge the gap from knowledge toq 1996 Academic Press, Inc. action’’ (p. 10). Two decades later, Nisbett and Ross were continuing to despair ‘‘our field’s inability toIt is always thus, impelled by a state of mind which is destined
not to last, that we make our irrevocable decisions. bridge the gap between cognition and behavior, a gap Marcel Proust that in our opinion is the most serious failing of mod-
Das ist eine Versuchung, sagte der Hofprediger und erlag ihr.1 ern cognitive psychology’’ (1980, p. 11). This essay is Bertolt Brecht, Mutter Courage
an attempt to construct the foundation for a bridge across the gap between perceived self-interest andThe ideas in this essay were stimulated by discussions with Drazen behavior. I argue that disjunctions between per-Prelec, and the formal analysis in section III is adapted from our joint
grant proposal. I thank Baruch Fischhoff, Chris Hsee, Helmut Junger- ceived self-interest and behavior result from the ac- mann, Daniel Kahneman, Gideon Keren, Sam Issacharoff, Graham tion of visceral factors such as the cravings associated Loomes, Daniel Nagin, Fritz Oser, and Peter Ubel for numerous helpful with drug addiction, drive states (e.g., hunger, thirst, discussions, suggestions, and comments. Address reprint requests to
and sexual desire), moods and emotions, and physicalGeorge Loewenstein, Department of Social and Decision Sciences, Car- pain. At sufficient levels of intensity, these, and mostnegie Mellon University, Pittsburgh, PA 15213-3890.
1 ‘‘This is a temptation, the court priest said, then succumbed.’’ other visceral factors, cause people to behave con-
2720749-5978/96 $18.00 Copyright q 1996 by Academic Press, Inc. All rights of reproduction in any form reserved.
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
273VISCERAL FACTORS
trary to their own long-term self-interest, often with Rational choice requires that visceral factors be taken into account. It makes good sense to eat whenfull awareness that they are doing so.
The defining characteristics of visceral factors are, first, hungry, to have sex when amorous, and to take pain killers when in pain. However, many classic patternsa direct hedonic impact, and second, an influence on the
relative desirability of different goods and actions. Hun- of self-destructive behavior, such as overeating, sex- ual misconduct, substance abuse, and crimes of pas-ger, for example, is an aversive sensation that affects the
desirability of eating. Anger is also typically unpleasant sion, seem to reflect an excessive influence of visceral factors on behavior. As the intensity of a specific vis-and increases one’s taste for various types of aggressive
actions. Physical pain is, needless to say, painful and ceral factor increases, its influence on behavior tends to increase and to change in a characteristic fashion.enhances the attractiveness of pain killers, food, and sex.
Although from a purely formal standpoint one could re- At low levels of intensity, people seem to be capable of dealing with visceral factors in a relatively optimalgard visceral factors as inputs into tastes, such an ap-
proach would obscure several crucial qualitative differ- fashion. For example, someone who is slightly sleepy might decide to leave work early or to forgo an eve-ences between visceral factors and tastes: ning’s planned entertainment so as to catch up on(1) Changes in visceral factors have direct hedonic sleep. There is nothing obviously self-destructiveconsequences, holding actual consumption constant. In about these decisions, even though they may notthat sense, visceral factors resemble consumption, not maximize ex post utility in every instance. Increasestastes. Whether I would be better off having one set in the intensity of visceral factors, however, oftenor preferences or another is an abstract philosophical produce clearly suboptimal patterns of behavior. Forquestion; whether I would be better off hungry or sati- example, the momentary painfulness of rising earlyated, angry or calm, in pain or pain-free, in each case produces ‘‘sleeping in’’ — a behavioral syndrome withholding consumption constant, is as obvious as whether wide-ranging negative consequences. It is at interme-I would prefer to consume more or less, holding tastes diate levels of intensity that one observes the classicand visceral factors constant. cases of impulsive behavior and efforts at self-con-(2) Changes in visceral factors are predictably corre- trol — e.g., placing the alarm clock across the roomlated with external circumstances (stimulation, depri- (Schelling, 1984). Finally, at even greater levels ofvation, and such) and do not imply a permanent change intensity, visceral factors can be so powerful as toin a person’s behavioral dispositions. In contrast, virtually preclude decision making. No one decideschanges in preferences are caused by slow experience to fall asleep at the wheel, but many people do.and reflection, are typically not anticipated, and do im-
The overriding of rational deliberation by the influ-ply a permanent change in behavior. ence of visceral factors is well illustrated by the behav-(3) Visceral factors typically change more rapidly ior of phobics who are typically perfectly aware thatthan tastes. Tastes also change, but tend to be stable the object of their fear is objectively nonthreatening,in the short run. but are prevented by their own fear from acting on this(4) Finally, tastes and visceral factors draw on dif- judgment (Epstein, 1994, p. 711). It can also be seenferent neurophysiological mechanisms. As Pribram in behaviors commonly associated with addiction, such(1984, p. 2) writes, ‘‘the core of the brain . . . uses as that of Charlie T, a former heroin addict whose urinechemical regulations to control body functions. The con- test showed that he had suddenly used heroin after afiguration of concentrations of these chemicals, al- long hiatus. Charlie was ‘‘overwhelmed by an irresist-though fluctuating around some set point, is suffi-
ciently stable over periods of time to constitute steady ible craving and . . . rushed out of his house to find ‘states.’ These states apparently are experienced as some heroin. . . . It was as though he were driven by hunger, thirst, sleepiness, elation, depression, effort, some external force he was powerless to resist, even comfort, and so on.’’ Their common neurochemical basis though he knew while it was happening that it was a may explain why so many behavior disorders associ- disastrous course of action for him’’ (Goldstein, 1994, ated with visceral factors — e.g., overeating, compulsive p. 220, emphasis added). Behavior at variance with shopping, phobias, and drug addictions — appear to be susceptible to moderation by a single drug: Fluoxetine
portant interdependencies between them. Tastes are importantly(Messiha, 1993). Tastes, in contrast to visceral factors, shaped by visceral factors. For example, one’s taste for barbecuedconsist of information stored in memory concerning the chicken may well underlie one’s visceral reaction to the smell of co-
relative desirability of different goods and activities.2 mingled charcoal, grease, and tomato sauce. At the same time, the visceral hunger produced by such smells, and the visceral pleasure
2 Although visceral factors are distinct from tastes in underlying produced by subsequent consumption, are likely to reinforce one’s preexisting taste for barbecued chicken.mechanisms and effects on well-being and behavior, there are im-
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
274 GEORGE LOEWENSTEIN
deliberation, however, is by no means confined to the tioning.4 Similarly, though by mutual agreement, labor negotiations are commonly structured to go ‘‘round therealm of the ‘‘abnormal.’’ Adam Smith, for example,
who is widely viewed as a proponent of enlightened clock’’ as the strike deadline approaches. Rarely is new information produced in these last sessions, nor isself-interest, described his own internal conflict — pre-
sumably in the face of sexual desire — as follows: there a discussion of technicalities of agreement. Per- haps, however, both sides recognize that mutual will-
At the very time of acting, at the moment in which passion ingness to make concessions will be enhanced when mounts the highest, he hesitates and trembles at the thought of sleep is the reward for speedy reconciliation. what he is about to do: he is secretly conscious to himself that
Decision theory, as it is currently practiced, makeshe is breaking through those measures of conduct which, in all no distinction between visceral factors and tastes andhis cool hours, he had resolved never to infringe, which he had
never seen infringed by others without the highest disapproba- thus does not recognize the special impact of visceral tion, and the infringement of which, his own mind forebodes, factors on behavior. It is best equipped to deal with must soon render him the object of the same disagreeable senti- ‘‘cool’’ or ‘‘dispassionate’’ settings in which there is typi- ments. (1892/1759, p. 227)
cally a very close connection between perceived self- interest and behavior. The decision-making paradigmSuccess, in many professions, is achieved through has much greater difficulty in providing an account ofa skillful manipulation of visceral factors. Automobile decisions occurring at the ‘‘hot’’ end of the continuumsalespersons, realtors, and other professionals who use defined by the intensity of visceral factors. The drive‘‘high pressure’’ sales tactics, for example, are skillful mechanism of Freudian and behavioristic psychologymanipulators of emotions. Con men are likewise expert provides a better account of behavior at the oppositeat rapidly invoking greed, pity, and other emotions that end of the same continuum. The decline of the behav-can eclipse deliberation and produce an override of nor- iorist paradigm in psychology can be attributed to itsmal behavioral restraints. Cults and cult-like groups failure to make sense of volitional, deliberative, behav-such as ‘‘EST’’ use food deprivation, forced inconti- ior. Does the decision-making paradigm face a similarnence, and various forms of social pressure in their fate if it fails to address the full range of visceral influ-efforts to recruit new members (Cinnamon & Farson, ences? My intent is to show that visceral influences on1979; Galanter, 1989). In all of these cases there is a behavior can, in fact, be expressed in decision-theoreticstrong emphasis on the importance of immediate ac- terms. Section II below addresses the question of whytion — presumably because influence peddlers recog- and how visceral factors create discrepancies betweennize that visceral factors tend to subside over time. The perceived self-interest and behavior. Section III enu-car or house one is considering will be ‘‘snapped up’’ if merates a series of propositions concerning the effectnot purchased immediately, and the one-time-only deal of visceral factors on behavior and perceptions, andon the stereo system will expire. The once-in-a-lifetime shows how these can be expressed in the verbal andopportunity for enrichment will be lost if one doesn’t mathematical language of decision-theory. Section IVentrust one’s bank card to the con artist, and there is discusses applications of the proposed theoretical per-an unexplained urgency to the insistence that one signs spective.up for EST in the introductory meeting rather than at
home after careful deliberation. II. VISCERAL FACTORS AND BEHAVIORTactics of this type are not, however, restricted to
those involved in the selling professions. Interrogators As visceral factors intensify, they focus attention and
use hunger, thirst, and sleep deprivation to extract con- motivation on activities and forms of consumption that
fessions. Like Esau, who sold his birthright for a mess are associated with the visceral factor — e.g., hunger
of pottage, prisoners may sacrifice years of freedom for draws attention and motivation to food. Non-associated
an extra hour of sleep or a glass of water. Lawyers use forms of consumption lose their value (Easterbrook,
a similar tactic when taking depositions.3 The early 1959). At sufficient levels of intensity, individuals will
stages of a deposition, when the witness is fresh, are sacrifice almost any quantity of goods not associated
used to elicit background information. Information that with the visceral factor for even a small amount of
is potentially damaging to the witness or the opposing associated goods, a pattern that is most dramatically
side is requested only after the witness begins to tire, lose concentration, and is more likely to make mistakes
4 Just as a skillful lawyers strategically manipulate the opposingor concessions just for the sake of ending the ques- side’s emotions, they must also work to counteract such influences on themselves and their clients. The adage ‘‘the lawyer who represents himself has a fool for a client’’ reflects the dangers to a lawyer of
3 Personal communication, Sam Issacharoff. excessive personal — i.e., emotional — involvement in a case.
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
275VISCERAL FACTORS
evident in the behavior of drug addicts. Frawley (1988, Evolution, however, has its limitations (Gould, 1992). The same visceral factors that serve the individ-p. 32) describes addicts as progressively ‘‘eliminating
behavior that interferes with or does not lead to drug ual’s interests effectively at moderate levels produce distinctly suboptimal patterns of behavior at higheror alcohol use . . . [which] leads to a kind of ‘tunnel
vision’ on the part of the addict.’’ Cocaine addicts, ac- levels. Extreme fear produces panic and immobiliza- tion rather than effective escape (Janis, 1967; Janis &cording to Gawin (1991, p. 1581), ‘‘report that virtually
all thoughts are focused on cocaine during binges; nour- Leventhal, 1967). Uncontrolled anger produces ineffec- tual, impulsive actions or the opposite, immobilization.ishment, sleep, money, loved ones, responsibility, and
survival lose all significance.’’ In economic parlance, Intense visceral factors not only undermine effective behavior, but produce extreme misery. This should notthe marginal rate of substitution between goods associ-
ated with the visceral factor and goods that are not so- surprise us; the ‘goal’ of evolution is reproduction, not happiness. If hunger ensures that an organism will eat,associated becomes infinitessimal.
Visceral factors also produce a second form of atten- the fact that it is an unpleasant sensation is immate- rial. As Damasio (1994, p. 264) argues, visceral factorstion-narrowing: a good-specific collapsing of one’s time-
perspective toward the present. A hungry person, for tend to be aversive because ‘‘suffering puts us on notice. example, is likely to make short-sighted tradeoffs be- Suffering offers us the best protection for survival, tween immediate and delayed food, even if tomorrow’s since it increases the probability that individuals will hunger promises to be as intense as today’s. This pres- heed pain signals and act to avert their source or cor- ent-orientation, however, applies only to goods that are rect their consequences.’’ associated with the visceral factor, and only to tradeoffs Although visceral factors should be and are taken between the present and some other point in time. A into account in decision making, they also influence hungry person would probably make the same choices behavior more directly. Hunger, thirst, sexual desire, as a non-hungry person between immediate and de- pain, and indeed virtually all visceral factors, can layed money (assuming that food cannot be purchased) influence behavior without conscious cognitive medi- or immediate and delayed sex. A hungry person might ation (Bolles, 1975). To illustrate this point, Pribram also make the same choices as a non-hungry person (1984) provides the vivid example of a brain surgery between food tomorrow versus food on the day after patient who ate ravenously with no subjective feeling tomorrow. of hunger:
Yet a third form of attention-narrowing involves the One patient who had gained more than one hundred pounds in
self versus others. Intense visceral factors tend to nar- the years since surgery was examined at lunch time. Was she row one’s focus inwardly — to undermine altruism. Peo- hungry? She answered, ‘‘No.’’ Would she like a piece of rare,
juicy steak? ‘‘No.’’ Would she like a piece of chocolate candy? Sheple who are hungry, in pain, angry, or craving drugs answered, ‘‘Umhumm,’’ but when no candy was offered she didtend to be selfish. As interrogators understand all too not pursue the matter. A few minutes later, when the examina-well, sleep deprivation, hunger, thirst, pain, and indeed tion was completed, the doors to the common room were opened
most visceral factors, can cause even the most strongly and she saw the other patients already seated at a long table willed individuals to ‘‘betray’’ comrades, friends and eating lunch. She rushed to the table, pushed the others aside, family (e.g., Biderman, 1960). and began to stuff food into her mouth with both hands. She
was immediately recalled to the examining room and the ques-The peremptory nature of immediate visceral factors tions about food were repeated. The same negative answers wereis generally adaptive. Visceral factors play an im- obtained again, even after they were pointedly contrasted with
portant role in regulating behavior, and can be ob- her recent behavior at the table. (p. 24).
served in a wide range of animals. Hunger signals the Further evidence for the direct impact of visceral fac-need for nutritional input, pain indicates the impinge- tors — without deliberative mediation — comes fromment of some type of potentially harmful environmen- neuropsychological research. This research shows, fortal factor, and emotions serve a range of interrupting, example, that brain lesions in the reward centers ofprioritizing, and energizing functions (Simon, 1967; the brain can produce a total lack of interest in eatingMandler, 1964; Pluchik, 1984; Frank, 1988). The ab- (Bolles, 1975). Electrical stimulation of the same areassence of even one of these signalling systems detracts can produce complex sequences of behavior withoutdramatically from an individual’s quality of life and conscious mediation (Gardner, 1992, p. 71). Many ofchances of survival. Although most people occasionally the sensory organs have direct nerve connections towish they could eschew pain, one only has to witness these pleasure/motivation centers, strongly hinting atthe playground behavior of children who are congeni- the possibility that sensory inputs can have a directtally incapable of experiencing pain (and to observe the influence on behavior. Electrical stimulation of theseperpetual vigilance of their parents) to abandon this
fantasy (Fields, 1987, pp. 2 – 4). same regions is so pleasurable that animals will self-
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
276 GEORGE LOEWENSTEIN
administer such stimulation in preference to food, wa- for many actions that are subjectively experienced as purposive by decision makers, electrical impulses asso-ter, and sex, and will do so until the point of collapse ciated with the action begin fractions of a second beforeand even death (Olds & Milner, 1954). Similarly self- any conscious awareness of the intention to act (Libet,destructive patterns of behavior are exhibited by both Gleason, Wright, & Pearl, 1983).animals and humans towards addictive substances,
The issue of cognitive versus visceral control of be-such as crack cocaine, which have a very similar effect havior remains unresolved, and some compromise posi-on the reward centers of the brain as electrical stimula- tion may well ultimately prevail. At present, however,tion (Pickens & Harris, 1968). It is difficult to imagine there is little evidence beyond fallible introspectionthat this type of behavior reflects the outcome of a ratio- supporting the standard decision-theoretic assumptionnal decision process, since the rather rapid conse- of complete volitional control of behavior.quence is to eliminate the capacity to experience plea-
sure altogether. Again, these findings suggest that there are certain types of influences or incentives that III. SEVEN PROPOSITIONS AND A
MATHEMATICAL REPRESENTATIONoperate independently of, and overwhelm, individual deliberation and volition.
Much is known, or at least can plausibly be inferredIn contrast to this relatively strong evidence that from available evidence, about the relationship be-visceral factors can influence behavior directly, there is tween deliberation and action under the influence ofonly weak evidence supporting the standard decision- visceral factors. The propositions enumerated belowtheoretic assumption that behavior follows automati- can be summarized simply: visceral factors operatingcally from deliberation. In fact, the standard decision- on us in the here and now have a disproportionatetheoretic assumption seems to be supported by little impact on our behavior. Visceral factors operating inmore than introspection. Most people experience their the past or future, or experienced by another individualown actions as resulting from decisions (Pettit, 1991), are, if anything, underweighted. Although these propo-or at least as deliberate. However, it is questionable sitions are simple enough to be stated in words, forwhether these introspections represent veridical re- the interested reader I also indicate how they could beports of underlying decision processes, or ex post ratio- expressed mathematically.nalizations of behavior. The limitation of verbal reports
To represent the influence of visceral factors on be-is well established (Nisbett & Wilson, 1977), as is the havior we need a representation of preferences thatfact that ‘‘implicit theories’’ powerfully influence one’s includes a new set of variables, ati , to represent howperception of the world (Bruner, 1957; Ross, 1989). Peo- the fluctuating levels of the visceral factors affect inter-ple process information in a hyper-Bayesian fashion, temporal utility:ignoring or down-playing evidence that is at variance
with their implicit theories while placing great weight on data that is supportive (Lord, Lepper & Ross, 1979). U Å Stu(xt1 , . . . , xtn , at1 , . . . , atm , t), (1) Trained to view behavior as the result of attribute- based decisions (Pettit, 1991; Christensen & Turner, where U is the total utility of an intertemporal con- 1993), most people in Western culture will almost inev- sumption plan, (xt1 , . . . , xtn) is the consumption vector itably interpret their own behavior accordingly. at time t, and a Å (at1 , . . . , atm) is the vector of visceral
Such a tendency to make retrospective sense of one’s factors at time t. In a given experiment, the a parame- own preferences and behavior can be seen in research ters will be operationally defined, e.g., as the hours of by Robert Zajonc and his colleagues on the ‘‘mere expo- food deprivation, the presence or absence of food stim- sure effect’’ (e.g., Zajonc, 1968). People are unaware of uli, and so on. We assume that the person knows the the effect of ‘‘mere’’ exposure on their preferences, but, values of x, a, and t when choosing between different when preferences are experimentally influenced consumption opportunities. through differential exposure, they readily generate at- Equation (1) is the most general temporally separa- tribute-based explanations for their own preferences ble model, and it allows for the value of any good or (Zajonc & Marcus, 1982). A subject might decide that activity to be affected by all visceral factors operating he likes polygon number 3, for example, not because he at the same point in time. In many instances, how- viewed it 12 times, but due to its geometric symmetry. ever, it is possible to partition visceral factors into Likewise, someone suffering from a tic that causes his subsets that influence only a single consumption hand to fly toward his head periodically will, over time, variable. In the simplest case, each consumption develop a head-itch that requires scratching (Brown, variable, xi , is influenced by at most one visceral fac-
tor, ai , as in Eq. (2).1988). Recent neuropsychological research shows that,
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
277VISCERAL FACTORS
value of a particular action or consumption alternative,U Å Stu(v1(xt1 , at1 , t), . . . , vn(xtn , atn , t)). (2) then proposition 1 implies that
In this equation, v1(xt1 , at1 , t) might be, say, the value If a* ú a ú a*i , and vd(x*, a*, 0)of consuming meal xt1 at time t relative to the present,
given that one’s hunger will be at level at1 at that time. Å vd(x, a, 0), then v(x*, a*, 0) ú v(x, a, 0). The separability structure in Eq. (2) implies that the ‘‘conditional’’ preference ordering of triples (xti , ati , t), This regularity was illustrated in the introduction with holding all else constant, is independent of the levels the example of sleepiness, which can be dealt with in of other consumption variables and visceral factors. a reasonable fashion at low levels, but at high levels Stable preferences across different types of consump- produces self-destructive patterns of behavior such as tion are captured by the function u(v1 , . . . , vn). The falling asleep at the wheel. A similar pattern of initially function tells us whether a person prefers dining out reasonable, but ultimately excessive, influence can be to dancing, for instance. The subordinate functions, vi , observed for virtually all visceral factors. Low levels of tell us how the value of particular dining opportunity fear may be dealt with in an optimal fashion (e.g., by hinges on what is offered (xit), the hunger level (ait), taking deliberate protective action), but higher levels and delay (t). Each of the vi functions is assumed to be of fear often produce panic or, perhaps worse, immobili- increasing in the first variable, decreasing in the third, zation (Janis, 1967). Likewise, low levels of anger can and possibly increasing or decreasing in the second. be factored into daily decision making in a reasonable Further, xi and ai will usually be complements, e.g., way, but high levels of anger often produce impulsive, hunger will enhance a solid meal, but hurt when no self-destructive, behavior. food is forthcoming. I also assume that xi and ai have
Proposition 2:natural zero levels. For xi , it is the status quo, or refer- Future visceral factors produce little discrepancy between theence consumption level (Tversky & Kahneman, 1991). value we plan to place on goods in the future and the value we
For ai , it is the level a*i such that v(0, a*i , t) Å 0. Intu- view as desirable. itively, the natural zero level of a visceral factor is the level at which, in the absence of the relevant form of That is, if a* ú a ú a*i , and vd(x*, a*, t) Å vd(x, a, t), consumption, the visceral factor neither contributes to then v(x*, a*, t) É v(x, a, t), for t ú 0. nor detracts from utility. When visceral factors are not having an immediate
influence on our behavior, but will be experienced in the future, we are free to give them the weight that wePropositions deem appropriate in decision making. Thus, we posi-
The observation that visceral factors influence the tion the alarm clock across the room to prevent sleeping desirability of goods and activities is hardly surprising. late only because we are not currently experiencing the To provide useful insights into behavior it is necessary pain of rising early. Likewise, we avoid buying sweets to specify the nature of this influence with the greatest when shopping after lunch because the evening’s crav- detail possible given the available evidence. The follow- ings, however predictable, have little reality to our cur- ing seven propositions, which are summarized in Table rent, unhungry selves. When the future becomes the 1 and discussed in detail below, encode observations present, however, and we actually experience the vis- concerning the influence of visceral factors on desired, ceral factor, its influence on our behavior is much predicted, recollected, and actual behavior. Although greater, as implied by proposition 1. all seven have some support from existing research, I A well-known study of pregnant women’s decisions refer to them as propositions to emphasize their tenta- concerning anesthesia illustrates the types of behav- tive status. ioral phenomena associated with proposition 2. Chris-
tensen-Szalanski (1984) asked expectant women to Proposition 1: make a non-binding decision about whether to use an- The discrepancy between the actual and desired value placed on
esthesia during childbirth; a majority stated a desirea particular good or activity increases with the intensity of the to eschew anesthesia. However, following the onset ofimmediate good-relevant visceral factor.5 labor, when they began to experience pain, most re-
If we define vd as the desired, as opposed to the actual, versed their decision. Consistent with proposition 2, the women were relatively cavalier with respect to their own future pain. Although Christensen-Szalanski5 By ‘‘actual value’’ I mean the value implied by the individual’s himself explained the reversals in terms of hyperbolicbehavior; by ‘‘desired value,’’ I mean the value that the individual
views as in his or her self-interest. discounting curves, such an account should predict that
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
278 GEORGE LOEWENSTEIN
TABLE 1
Propositions Concerning the Actual, Desired, Predicted, and Recollected Influence of Visceral Factors on Behavior
Proposition Description
1 The discrepancy between the actual and desired value placed on a particular good or activity increases with the intensity of the immediate good-relevant visceral factor.
2 Future visceral factors produce little discrepancy between the value we plan to place on goods in the future and the value we view as desirable.
3 Increasing the level of an immediate and delayed visceral factor simultaneously enhances the actual valuation of immediate relative to delayed consumption of the associated good.
4 Currently experienced visceral factors have a mild effect on decisions for the future, even when those factors will not be operative in the future.
5 People underestimate the impact of visceral factors on their own future behavior. 6 As time passes, people forget the degree of influence that visceral factors had on their own past behavior. As a
result, past behavior that occurred under the influence of visceral factors will increasingly be forgotten, or will seem perplexing to the individual.
7 The first six propositions apply to interpersonal as well as intrapersonal comparisons, where other people play the same role vis a vis the self as the delayed self plays relative to the current self:
i. We tend to become less altruistic than we would like to be when visceral factors intensify. ii. When making decisions for another person, we tend to ignore or give little weight to visceral factors
they are experiencing iii. Increasing the intensity of a visceral factor for ourselves and another person in parallel leads to a
decline in altruism. iv. When we experience a particular visceral factor, we tend to imagine others experiencing it as well,
regardless of whether they actually are. v. & vi. People underestimate the impact of visceral factors on other people’s behavior.
only reason to stop ‘‘using’’ is to avoid negative consequencesat least some reversals would occur prior to the onset that accompany continuing usage. (p. 152)of labor, but none did. Moreover, the reversal of prefer-
ence was observed not only for women giving birth for In a similar vein, Osiatynski refers to the tendency to the first time, but also those who had previously experi- underestimate the power of alcohol addiction: ‘‘After enced the pain of childbirth; experience does not seem hitting bottom and achieving sobriety, many alcoholics to go very far in terms of enhancing one’s appreciation must get drunk again, often not once but a few times, for future pain. in order to come to believe and never forget about their
A similarly underappreciation of the impact of future powerlessness’’ (1992, p. 128). Osiatynsi argues that a visceral states — again by people with considerable ex- major task of relapse prevention is to sustain the ex- perience — can be seen in the relapse behavior of ad- addict’s appreciation for the force of craving and the dicts who, after achieving a period of abstinence, be- miseries of addiction; alcoholics anonymous serves this lieve they can indulge in low level consumption without function by exposing abstinent alcoholics to a continual relapsing. Underestimating the impact of the craving stream of new inductees who provide graphic reports that even small amounts of consumption can produce of their own current or recent miseries.6
(Gardner & Lowinson, 1993), such addicts typically Impulsivityfind themselves rapidly resuming their original ad-
dictive pattern of consumption (Stewart & Wise, 1992). The disproportionate response to immediately opera- As Seeburger (1993) comments: tive visceral factors expressed by proposition 1, and the
tendency to give little weight to delayed visceral factors Any addict can tell us how long such negative motivation [to
expressed by proposition 2, have important implica-stay off the drug] lasts. It lasts as long as the memory of the tions for intertemporal choice.7 Together they point toundesirable consequences stays strong. But the more successful
one is at avoiding an addictive practice on the grounds of such a novel account of impulsivity — an alternative to the motivation, the less strong does that very memory become. Be- currently dominant account which is based on non-ex- fore long, the memory of the pain that one brought on oneself ponential time discounting. through the addiction begins to pale in comparison to the antici- pation of the satisfaction that would immediately attend relapse
6 Personal communication.into the addiction. Sometimes in AA it is said that the farther away one is from one’s last drink, the closer one is to the next 7 For a preliminary rendition of this perspective, see Hoch and
Loewenstein (1991).one. That is surely true for alcoholics and all other addicts whose
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
279VISCERAL FACTORS
In a seminal article, R. H. Strotz (1955) showed that The account of impulsivity embodied in propositions 1 and 2 is consistent with the observed differences ina discounted utility maximizer who does not discount impulsivity across goods and situations. It views impul-at a constant rate will systematically depart from his sivity as resulting not from the disproportionate attrac-own prior consumption plans. When the deviation from tiveness of immediately available rewards but from theconstant discounting involves higher proportionate dis- disproportionate effect of visceral factors on the desir-counting of shorter time delays than of long ones, this ability of immediate consumption. It predicts, there-‘‘time inconsistency’’ takes the form of temporally myo- fore, that impulsive behavior will tend to occur whenpic or impulsive behavior: spending in the present but visceral factors such as hunger, thirst, physical pain,vowing to save in the future, binge-eating in the pres- sexual desire, or emotions are intense. In combination,ent while planning future diets, or resolving to quit propositions 1 and 2 imply that people will give muchsmoking, but not until tomorrow. A standard non-expo- greater weight to immediately experienced visceral fac-nential discounting formulation that predicts impul- tors than to delayed visceral factors. Thus, accordingsive behavior is U Å u(x0) / gdu(x1) / gd2u(x2), where to proposition 2, the fact that I will be hungry (andd is the conventional exponential discount factor and dying to eat dessert), in pain (and longing for pain kill-g(õ1) is a special discount factor applying to all periods ers), or sexually deprived in the future has little mean-other than the immediate present (see Elster, 1977; ing to me in the present. If food, pain killers, or sexAkerlof, 1991). The conventional, i.e., constant dis- have undesirable consequences I will plan to desistcounting, approach is identical, except that g is as- from these behaviors. When these visceral factorssumed to equal unity. A person who maximizes a func- arise, however, and increase my momentary valuationtion of this type will choose a larger reward x* at time of these activities, proposition 1 implies that I will devi-2 over a smaller reward x at time 1 if du(x*) ú u(x), but ate from my prior plans. In fact, neither propositionwill opt for the smaller, more immediate reward if the 1 nor 2 are necessary conditions for this account ofchoice is between immediate consumption or consump- impulsivity; what is required is a somewhat weakertion at time 1 if gdu(x*) õ u(x). condition which can be expressed as a third proposition.The non-exponential discounting perspective has
been bolstered by findings from hundreds of experi- Proposition 3: ments showing that humans and a wide range of other Increasing the level of an immediate and delayed visceral factor
simultaneously enhances the actual valuation of immediate rela-animals, display hyperbolic discount functions of the tive to delayed consumption of the associated good.type predicted to produce impulsive behavior (see, e.g.,
Chung & Herrnstein, 1967; Mazur, 1987). Many experi- That is, if a* ú a and v(x, a, 0) Å v(x*, a, t), then v(x, ments with animals, and a small number with humans, a*, 0) ú v(x*, a*, t). Whereas propositions 1 and 2 deal have also demonstrated the types of temporally based with the effect of visceral factors on the relationship preference reversals that are implied by hyperbolic dis- between actual and desired behavior, proposition 3 counting. Nevertheless, the non-exponential dis- makes no reference to desired behavior and refers only counting perspective has at least two significant limita- to the impact of visceral factors on time preference. The tions as a general theory of impulsivity. absence of the subjective concept of desired behavior
First, it does not shed light on why certain types renders proposition 3 especially amenable to empirical of consumption are commonly associated with impul- investigation. sivity while others are not. People commonly display Like the hyperbolic discounting perspective, the vis- impulsive behavior while under the influence of vis- ceral factor perspective predicts that impulsivity will ceral factors such as hunger, thirst, or sexual desire often be associated with short time delays to consump- or emotional states such as anger or fear. The hyper- tion; however, it provides a different rationale for this bolic discounting perspective has difficulty account- prediction and does not predict that short time delays ing for such situation- and reward-specific variations will always produce impulsive behavior. According to in impulsivity. the hyperbolic discounting perspective, desirability in-
Second, the hyperbolic discounting perspective can- creases automatically when rewards become immi- not explain why many situational features other than nently available. The visceral factor perspective, in con- time delay — for example, physical proximity and sen- trast, assumes that immediate availability produces sory contact with a desired object — are commonly asso- impulsivity only when physical proximity elicits an ap- ciated with impulsive behavior. For example, it is diffi- petitive response (influences an a). Many visceral fac- cult to explain the impulsive behavior evoked by cookie tors, such as hunger and sexual desire, are powerfully shops that vent baking smells into shopping malls in influenced by temporal proximity. Neurochemical re-
search on animals shows that the expectation of anterms of hyperbolic discounting.
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
280 GEORGE LOEWENSTEIN
imminent reward produces an aversive dopaminic ing times. Apparently the photograph provided a ‘‘pic- ture’’ of the benefits of waiting without increasing thestate in the brain that is analogous to the impact of child’s level of acute hunger or desire. Likewise, andfood expectation on hunger (Gratton & Wise, 1994). explicable in similar terms, instructing children to ig-That is, the mere expectation of an imminent reward nore the candies or to cognitively restructure themseems to trigger appetite-like mechanisms at the most (e.g., by thinking of chocolate bars as little brown logs)basic level of the brain’s reward system. The account also increased waiting times.of impulsivity proposed here, therefore, predicts that
short time delays will elicit impulsivity only when they Vividnessproduce such an appetitive, or other type of visceral,
response. The notion that various dimensions of proximity —
Short time delays, however, are only one factor that temporal, physical, and sensory — can elicit visceral in-
can produce such a visceral response. Other forms of fluences that change behavior also provides a some-
proximity, such as physical closeness or sensory contact what different interpretation of the often noted effect of
(the sight, smell, sound, or feeling of a desired object) vividness. Vividness has a powerful impact on behavior
can elicit visceral cravings. Indeed, as the literature on that is difficult to reconcile with the standard decision
conditioned craving in animals shows, almost any cue model. Sweepstakes advertise concrete grand prizes
associated with a reward — e.g., time of day, the color of such as luxury cars or vacations, even though any nor-
a room, or certain sounds — can produce an appetitive mative model would predict that the monetary equiva-
response (Siegel, 1979). Perhaps the strongest cue of lent of the prize should have higher value to most indi-
all, however, is a small taste, referred to as a ‘‘priming viduals. When Rock Hudson and Magic Johnson were
dose’’ in the neuropharmacological literature on drug diagnosed with AIDS, concern for the disease skyrock-
addiction (Gardner & Lowinson, 1993). eted (Loewenstein & Mather, 1990). Well-publicized in-
Much of the seminal research of Walter Mischel and cidents of ‘‘sudden acceleration’’ and terrorist attacks associates (summarized in Mischel 1974; Mischel, at airports in Europe squelched Audi sales and travel Shoda, & Yuichi, 1992) can be interpreted as demon- abroad by Americans, despite the comparative safety strating the impact of visceral factors on impulsivity. of Audis and foreign travel. Behavioral decision re- Mischel’s research focused on the determinants of de- searchers have acknowledged the impact of vividness lay of gratification in children and was the first to raise (Tversky & Kahneman, 1973; Nisbett & Ross, 1980), the problem of intraindividual variability in intertem- but have argued that vividness affects decision making poral choice. In a series of experiments, children were via its influence on subjective probability. Vividness is placed in a room by themselves and taught that they assumed to affect the ease with which past instances could summon the experimenter by ringing a bell. The of the outcome can be remembered or future instances children would then be shown a superior and inferior imagined, producing an exaggeration of the outcome’s prize and told that they would receive the superior subjective probability via the ‘‘availability heuristic.’’ prize if they could wait successfully for the experi- Vividness, however, has a second, possibly more im- menter to return. portant, consequence. Immediate emotions arising
One major finding was that children found it harder from future events are inevitably linked to some mental to wait for the delayed reward if they were made to image or representation of those events. There is con- wait in the presence of either one of the reward objects siderable research demonstrating that the more vivid (the immediate inferior or delayed superior). The fact such images are, and the greater detail with which they that the presence of either reward had this effect is are recalled, the greater will be the emotional response significant, because conventional analysis of intertem- (e.g., Miller et al., 1987). Hence, vividness may operate poral choice, including the hyperbolic discounting per- in part by intensifying immediate emotions associated spective, would predict that children would be more with thinking about the outcome rather than (or in likely to wait in the presence of the delayed reward. addition to) increasing the subjective likelihood of the The visceral factor perspective offers a ready explana- outcome. tion for this pattern, since the sight, smell, and physical Many phenomena which have previously been attrib- proximity of either reward would be likely to increase uted to availability effects on subjective probability the child’s level of hunger and desire. could easily be reinterpreted in these terms. It has been
Other findings from Mischel’s research are also con- shown, for example, that earthquake insurance pur- sistent with a visceral factor account of impulsivity. For chases rise after earthquakes when, if anything, the example, showing children a photograph of the delayed objective probability is probably at a low-point but anx-
iety about these hazards is at a peak (Palm, Hodgson,reward, rather than the reward itself, increased wait-
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
281VISCERAL FACTORS
Blanchard, & Lyons, 1990). Similarly, purchases of audience to stick 50 – 60 pins into him up to their heads, then would himself pull them out one by one (Morris,flood and earthquake insurance are influenced more by
whether friends have experienced the event than by the 1991). By Morris’ description, ‘‘it is clear that Gibson’s audience, no doubt reflecting a general human re-experience of one’s immediate neighbors, even though
neighbors’ experiences would seem to provide a better sponse, found themselves incapable of imagining a truly pain-free existence. They instinctively suppliedguide to one’s own probability of experiencing a flood
or earthquake (Kunreuther et al., 1978). The large in- the pain he did not feel’’ (p. 13). crease in the number of women seeking breast exams Proposition 5: following the highly publicized mastectomies of Hope People underestimate the impact of visceral factors on their own Rockefeller and Betty Ford, the tendency for doctors future behavior. whose specialties are near the lung to stop smoking,
Let v̂ represent the individual’s prediction at time t õ and each of the examples of vividness listed earlier
0 of the value she will place on consumption at time 0 could also plausibly be attributed to emotion effects
(when a visceral factor will be operative). Proposition rather than to changes in subjective probabilities. Most
5 implies that if a* ú a ú a*i , and v̂(x*, a*, 0) Å v̂(x, a,doctors have a clear understanding of the dangers of 0), then v(x*, a*, 0) ú v(x, a, 0).smoking, but daily confrontation with blackened lungs
Proposition 5 is similar to proposition 2 except thatundoubtedly increases the frequency and intensity of it refers to predictions of future behavior rather thannegative emotions associated with smoking. to decisions applying to the future. It implies that we
Proposition 4: underestimate the influence of future visceral factors Currently experienced visceral factors can have a mild effect on on our behavior, whereas proposition 2 implies that we decisions for the future, even when those factors will not be give future visceral factors little weight when making operative in the future.
decisions for the future. Although closely related, the two phenomena have somewhat opposite implicationsProposition 4 is probably a minor effect relative to
the other six discussed here, and it cannot be expressed for behavior; the failure to appreciate future visceral factors (as implied by proposition 2) increases our like-in conditions pertaining to Eq. (2), which assumes that
the value of consumption is influenced only by visceral lihood of binding our own future behavior — thus con- tributing to far-sighted decision making. For example,factors operating at the same point in time. To express
proposition 4 mathematically we could allow visceral showing little sensitivity to tomorrow morning’s self, we experience no qualms in placing the alarm clockfactors operating in the present to influence the value
of consumption at other points in time — e.g., vi(xti , ati , across the room. The failure to predict our own future behavior (as implied by condition 5), however, de-t, a0i). Proposition 4 would then imply that if a*oi ú a0i
and vi(xti , ati , t, a0i) Å vj(xtj , atj , t, a0j), then vi(xti , ati , creases the likelihood that we will take such actions, even when they are necessary. Failing to predict thet, a*oi) § vj(xtj , atj , t, a0j).
The classic illustration of proposition 4 is the ten- next morning’s pain of awakening, we may underesti- mate the necessity of placing the alarm clock on thedency to buy more groceries when shopping on an
empty stomach (Nisbett & Kanouse, 1968). Similarly, other side of the room. The difficulty of predicting the influence of futurewhen sick we are likely to overreact by cancelling ap-
pointments later in the week, only to find ourselves visceral factors on our behavior results partly from the fact that visceral factors are themselves difficult to pre-recovered on the following day. It also seems likely
that an aggrieved person would decide to take delayed dict. The strength of visceral factors depends on a wide range of influences. Drive states such as sexual desirerevenge if immediate revenge were not an option, even
if she knew intellectually that her anger was likely to and hunger depend on how recently the drive was satis- fied and on the presence of arousing stimuli such as‘‘blow over.’’
The same failure of perspective taking can be ob- potential sexual objects or the proximity of food. Moods and emotions depend on the interaction of situationalserved in the interpersonal realm. For example, it is
difficult for a parent, who feels hot from carrying a factors and construal processes and on internal psycho- biological factors. Physical pain and pleasure often de-baby, to recognize that his baby might not be as hot.
Similarly, it is difficult not to empathize with a pend on sensory stimulation, although construal pro- cesses also play an important role (Chapman, 1994).wounded person even when they report feeling no pain.
The latter phenomenon is illustrated vividly by the Because these underlying factors are themselves often erratic, predicting changes in visceral factors is com-case of Edward Gibson, the ‘‘human pincushion.’’ A
Vaudeville performer who experienced no pain, Gibson mensurately difficult. Even when visceral factors change in a regular fash-would walk onto the stage and ask a man from the
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
282 GEORGE LOEWENSTEIN
ion, however, people will not be able to predict such suited to storing information about visceral sensations. Recall of visual images actually activates many of thechange if they lack a theory of how they change over
time. Thus, Loewenstein and Adler (1995) demon- brain systems that are involved in visual perception (Kosslyn et al., 1993). Thus, it appears that to imaginestrated that people are unable to predict that owner-
ship will evoke attachment to objects and aversion to a visual scene is, in a very real sense, to ‘‘see’’ the scene again, albeit in distorted, incomplete, and less vividgiving them up, presumably because they, like social
scientists until recently, are unaware of the endow- form. The same probably applies to memory for music and words; one can render a tune in one’s head, orment effect. They elicited selling prices from subjects
actually endowed with an object and others who were articulate a word, without producing any externally audible sound.told they had a 50% chance of getting the object. Selling
prices were substantially higher for the former group, Except under exceptional circumstances,8 memory for pain, and probably other visceral factors, appearsand the valuations of subjects who were not sure of
getting the object were indistinguishable from the buy- to be qualitatively different from other forms of mem- ory. As Morley (1993) observes in an insightful paper,ing prices of subjects who did not have the object.
Moreover, even in the many cases when we can pre- we can easily recognize pain, but few can recall any of these sensations at will, at least in the sense of reexpe-dict the intensity of a particular visceral factor rela-
tively accurately, we may still have difficulty in pre- riencing them at any meaningful level. Morley distin- guishes between three possible variants of memory fordicting its impact on our own future behavior. It is one
thing to be intellectually aware that one will be hungry pain: (1) sensory reexperiencing of the pain; (2) remem- bering the sensory, intensity, and affective qualities ofor cold at a certain point in the future and another to
truly appreciate the impact of that hunger or cold on the pain without re-experiencing it; and (3) remember- ing the circumstances in which the pain was experi-one’s own future behavior. If a teenager tries crack once
for the experience, how difficult will he or she find it to enced. Most studies of memory for pain have focussed on the second variant and have obtained mixed results.desist from trying it again? How strong will a smoker’s
desire to smoke be if she goes to a bar where others For example, several studies have examined the accu- racy of women’s memory of the pain of childbirth —are smoking, or the ex-alcoholic’s desire for a drink if
he attends the annual Christmas party at his place of most employing a so-called visual analog scale (basi- cally a mark made on a thermometer scale) (e.g., Rofé &work? Proposition 5 implies that people who are not
experiencing these visceral factors will underestimate Algom, 1985; Norvell, Gaston-Johansson, & Fridh, 1987). These have been about evenly split in their con-their impact on their own future behavior.
The difficulty of anticipating the effect of future vis- clusions, with about half finding accurate recall of pain (or even slight retrospective exaggeration) and theceral factors on one’s own behavior is also illustrated by
a study in which subjects were informed of the Milgram other half finding significant, and in some cases quite substantial, under-remembering of pain.shock experiment findings and were asked to guess
what they personally would have done if they had been Morley himself (1993) conducted a study in which subjects completed a two-part survey on pain memo-subjects in the experiment. Most subjects in the piggy-
back study did not think that they themselves would ries. In the first part they were asked to recall a pain event and in the second they were asked questions de-have succumbed to the pressure to shock. Despite their
awareness that a substantial majority of subjects deliv- signed to measure the extent of the three variants of ered what they believed were powerful shocks, subjects underestimated the likely effect on their own behavior 8 Traumatic injury may be such a case. Katz and Melzack (1990)
argue, based on research on amputees experiencing the ‘‘phantomof being exposed to the authoritative and relentless limb’’ phenomenon, that amputees store pain memories in a ‘‘neuro-pressure of the experimenter. matrix’’ such that they can be retrieved and veridically reexperi- enced: ‘‘The results of the present study suggest that the somatosen-Proposition 6: sory memories described here are not merely images or cognitiveAs time passes, people forget the degree of influence that visceral recollections (although obviously a cognitive component is involved);factors had on their own past behavior. As a result, past behavior they are direct experiences of pain (and other sensations) that resem-that occurred under the influence of visceral factors will seem ble an earlier pain in location and quality’’ (p. 333). They summarizeincreasingly perplexing to the individual. different past studies of phantom limb pain in which 46, 79, 50, 17.5, 37.5, and 12.5% of patients who had lost limbs reported that theIf we define vr as the individual’s recollection at time t pain mimicked the original pain. There are problems with this re-ú 0 of his own past utility, then, if a* ú a ú a*i , and search, most notably the retrospective methodology which introduces
v(x*, a*, 0) Å v(x, a, 0), then vr(x*, a*, 0) õ vr(x, a, 0). the possibility of recall bias. However, at a minimum, the phantom Human memory is well suited to remembering visual limb research suggests that some people in some situations may, in
fact, be capable of remembering pain.images, words, and semantic meaning, but seems ill-
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
283VISCERAL FACTORS
pain memory dimensions. When asked questions about the reality (e.g., Linton, 1991; Rachman & Arntz, 1991). For such events there is evidence that what people re-the second variant type of pain memory, 59% were able
to recall at least some aspect of the pain sensation, member is what they expected to experience before- hand, rather than what they actually experiencedwhile the remaining 41% reported that they had no
recall of the pain sensation at all and were thus unable (Kent, 1985). A similar pattern holds for emotions. Some emo-to rate the vividness of their pain experience. For exam-
ple, one subject reported ‘‘I remember the pain getting tions are associated with straightforward cognitions. For example, anger may arise from a perceived in-worse and worse, but I can’t remember what the pain
felt like at all.’’ Not a single subject reported actually sult, shame or embarrassment from a faux pas. To the extent that the insult or faux pas can be conjuredreexperiencing the pain — i.e., Morley’s first variant of
pain memory. Consistent with these results, Strong- up in the mind, one can reproduce the emotion at any time, not just at the time when the instigatingman and Kemp (1991) found that spontaneous accounts
of pain tended to fit Morley’s first variant of pain mem- incident occurs (see, Strack, Schwarz & Gschnei- dinger, 1985, p. 1464).10 Thus, as for pain, the abilityory — remembering the circumstances in which the
pain was experienced. Their subjects were given a list to imagine the impact of future emotions depends on the concreteness and imaginability of the instigatingof 12 emotions and were asked for each to remember
a time they had experienced the emotion. They found stimuli. Moods or feeling states that have no obvious object, such as sadness or depression, by this reason-that, ‘‘overwhelmingly, the descriptions were of ‘objec-
tive’ details of the events rather than of the feelings of ing, will be especially prone to anticipatory underes- timation, as will pains and discomforts that are notthe respondents’’ (p. 195).
Scarry (1985, p. 15) notes a similar phenomenon associated with vivid images. The latter observation may help to explain an obser-when it comes to descriptions of pain; these rarely de-
scribe the pain itself, but typically focus either on the vation made by Irena Scherbakowa (personal commu- nication), on the basis of hundreds of interviews con-external agent of pain (e.g., ‘‘it feels as though a ham-
mer is coming down on my spine’’) or on the objective ducted with victims of Stalin’s terror. She noted that people who had ‘‘betrayed’’ friends or family, or con-bodily damage associated with the pain (‘‘it feels as if
my arm is broken at each joint and the jagged ends fessed to crimes they didn’t commit when they were tortured by such methods as being forced to stand inare sticking through the skin’’). Fienberg, Loftus, and
Tanur (1985, p. 592) reached virtually the same conclu- one position for hours, or prevented from sleeping, may have been particularly haunted by the memory yearssion in their review of the literature on memory for
pain which concluded with the question: ‘‘Is it pain that later because it was difficult to understand, in retro- spect, why they had succumbed to such seeminglypeople recall or is it really the events such as injuries
and severe illnesses?’’ ‘‘mild’’ methods. A similar observation was made by Biderman (1960) in his analysis of the retrospectiveWhether people can remember the sensory, inten-
sity, and affective qualities of a pain (Morley’s second reports of 220 repatriated U.S. Air Force prisoners cap- tured during the Korean war. According to Biderman,variant), therefore, or only the events that produced
the pain, the evidence is strong that most people cannot ‘‘the failure of the prisoner to recognize the sources of the compulsion he experiences in interrogation intensi-remember pain in the sense of reexperiencing it in
imagination (Morley’s first variant). We can recognize fies their effects, particularly the disabling effects of guilt reactions’’ (p. 145).pain all too effortlessly when it is experienced, but only
in a limited number of cases actually call it to mind Limitations in the memory for visceral sensations may also help to explain the disappointing results thatspontaneously — i.e., recall them — in the same way
that we can recall words or visual images.9 have been obtained by interventions designed to alter behavior by invoking fear. In some such efforts, suchThere may be certain types of visceral sensations,
however, which, if not remembered in Morley’s third as trying to ‘‘scare-straight’’ at-risk youths by exposing them to life in a maximum security penitentiary, thesense, at least evoke arousal upon recall. For pain, this
is true of those for which the pain-causing event can effect seems to have been opposite to what was in- tended (Finckenauer, 1982; Lewis, 1983). The standardbe imagined vividly. Highly imaginable events such as
dentist visits, cuts and wounds, and bone breakage pro- explanation for such an effect is that the fear communi- cation produced a defensive compensatory response.duce immediate anxiety and dread, to the point where
the recollection of the event may actually be worse than Perhaps, however, the paradoxical effect resulted from
9 Deleted in proof. 10 Jon Elster brought this point to my attention.
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
284 GEORGE LOEWENSTEIN
the weakness of the evoked response to the memory. If to proposition 3 states that the weight one places on oneself relative to other persons who are experiencingthinking about incarceration fails to evoke affect, even
after touring the facility, perhaps the youths in ques- equivalent levels of a visceral factor increases as the common level of the visceral factor intensifies. Hun-tion conclude that ‘‘I’ve experienced the worst, and it
must not be that bad since thinking about it leaves me ger, thirst, pain, and fear are all powerful antidotes to altruism (Loewenstein, forthcoming a).cold.’’ This conjecture is consistent with research on
people’s response to minimally, moderately, and Proposition 4 applied to the interpersonal domain implies that people who are themselves experiencing astrongly fear-arousing lectures about dental hygiene
(Janis & Feshbach, 1953). Immediately following the visceral factor will be more empathic toward, and more accurate predictors of, others who are experiencing thecommunication there appeared to be a monotonic rela-
tionship between fear intensity and vigilance; however, same visceral factor. One summer, for example, a friend mentioned his back problems to me. I responded1 week later the effect of the lectures on behavior was
inversely related to fear. sympathetically, but his pain had little reality until, when working in the garden one day, I suddenly feltIn sum, with certain important exceptions, it appears
that people can remember visceral sensations at a cog- something ‘‘give’’ in my back. My virtually instant reac- tion was to think of him and to feel deeply for the firstnitive level, but cannot reproduce them, even at dimin-
ished levels of intensity. It seems that the human brain time what he must have been experiencing all along. Despite such occasional examples of ‘‘priming,’’ how-is not well equipped for storing information about pain,
emotions, or other types of visceral influences, in the ever, in which one’s own weak experience of a visceral factor allow us to empathize with another person’ssame way that visual, verbal, and semantic informa-
tion is stored. We can recognize visceral sensations of- stronger one, in general, there seems to be an empathic gulf when it comes to appreciating another person’sten too effortlessly when they occur, but only in a lim-
ited number of cases actually call them to mind sponta- pain, hunger, fear, etc. As Elaine Scarry writes with respect to pain,neously — i.e., recall them — in the same way that we
can recall words or visual images. Unable to recall vis- When one speaks about ‘‘one’s own physical pain’’ and about ceral sensations as we can recall other types of informa- ‘‘another person’s physical pain,’’ one might almost appear to
be speaking about two wholly distinct orders of events. For thetion, their power over our behavior is difficult to make person whose pain it is, it is ‘‘effortlessly’’ grasped (that is, evensense of retrospectively or to anticipate prospectively. with the most heroic effort it cannot not be grasped); while for the person outside the sufferer’s body, what is ‘‘effortless’’ isProposition 7: not grasping it (it is easy to remain wholly unaware of its exis-Each of the first six propositions apply to interpersonal as well tence; even with effort, one may remain in doubt about its exis-as intrapersonal comparisons, where other people play the same tence or may retain the astonishing freedom of denying its exis-role visavis the self as the delayed self plays relative to the tence; and, finally, if with the best effort of sustained attentioncurrent self. one successfully apprehends it, the aversiveness of the ‘‘it’’ one apprehends will only be a shadowy fraction of the actual ‘it’).Analogous to proposition 1, actual altruism tends to (1985, p. 4).decline relative to desired altruism as visceral factors
intensify. A friend related to me the frenzied strug- Scarry argues that pain, uniquely, possesses such an empathic gulf, and attributes it to the poverty of lan-gles between passengers that occurred on a transat-
lantic flight when the plane suddenly dived and only guage when it comes to expressing pain. While agree- ing with her that such a gulf exists, I think it appliesabout half the oxygen masks dropped. Although fear
caused people to become self-centered, it seems likely to a much wider range of feelings than pain, doubt it arises from limitations of linguistic expression, andthat even as they grasped for their neighbor’s child’s
mask, they knew that they were violating their own also believe that virtually the same gulf exists when it comes to remembering or anticipating one’s own painmoral codes. The self-focusing effects of visceral fac-
tors is not surprising given the prioritizing and moti- and other visceral factors. Regardless of the source of such an empathic gulf, its existence implies that, analo-vating role that visceral factors play in human and
nonhuman behavior. Analogous to proposition 2, gous to proposition 5, people will have difficulty pre- dicting the behavior of other people who are experienc-when making decisions for others, we are likely to
ignore or radically underweight the impact of visceral ing intense visceral factors. Just as people underesti- mated the likelihood that they themselves would havefactors on them. Few of the classic tragedies (e.g., Eve
and the apple; Macbeth) would have happened if the conformed to the modal pattern of behavior in the Mil- gram experiment, for example, they also underesti-protagonists had turned over decision-making power
to a disinterested party. Combining both of these mated the likelihood that other, superficially described, persons would do so (Nisbett & Ross, 1980). Finally,analogous propositions, the interpersonal equivalent
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
285VISCERAL FACTORS
analogous to proposition 6, the behavior of other people despite the opportunities for data fitting inherent in the typical retrospective design. Even when applied toacting under the influence of visceral factors will seem
as incomprehensible as one’s own past visceral-factor- gambling — an activity which serves as the central met- aphor for the decision making perspective — decisioninfluenced behavior.
Most of the propositions just enumerated, including models have been largely unable to account for the ‘‘stylized’’ facts of aggregate behavior, let alone to pre-the 7th, are illustrated in William Styron’s autobio-
graphical treatise on depression. Depression fits the dict the behavior of individuals. Is it possible that part of the poor fit problem results from the decision makingdefinition of a visceral factor since it has a direct impact
on well-being and also influences the relative desirabil- paradigm’s failure to take account of visceral factors? In this section I discuss a variety of patterns and do-ity of different activities. Proposition 1 (the excessive
influence of immediately operative visceral factors) is mains of behavior in which I believe that visceral fac- tors are likely to play an especially prominent role.illustrated by the fact that while he was depressed
Styron experienced an almost overwhelming desire to commit suicide, but recognized that this was not in his Drug Addiction self-interest. This latter awareness induced him to seek
In the introduction of Addiction, Avrum Goldsteinpsychiatric help. Proposition 2 (the underweighting of expresses the central paradox of addiction as follows:future visceral factors), proposition 5 (underestimation
of the impact of future visceral factors), and proposition If you know that a certain addictive drug may give you tempo- rary pleasure but will, in the long run, kill you, damage your6 (the minimization in memory of the impact of past health seriously, cause harm to others, and bring you into con-visceral factors) are also all vividly described in the flict with the law, the rational response would be to avoid thatbook. When Styron was not feeling depressed, he re- drug. Why then, do we have a drug addiction problem at all? In
ports, depression had little reality to him; indeed, writ- our information-rich society, no addict can claim ignorance of ing the book was his attempt to come to terms with the consequences. this lack of intrapersonal empathy. Proposition 4 (the
Several different solutions to this riddle have been pro-projection of currently experienced visceral factors onto posed. Becker and Murphy (1988), for example, arguethe future) is well illustrated by the feeling he reports, that the addict begins taking the addictive substancewhile depressed, that the depression will never end — with a realistic anticipation of the consequences. Suchall the while recognizing intellectually that this is an account is unsatisfactory not only because it fails toprobably false. Finally proposition 7 (the analog be- fit the facts (e.g., it implies incorrectly that addicts willtween intra- and interpersonal empathy vis à vis the buy in bulk to save time and money in satisfying theireffect of visceral factors) is amply illustrated both from anticipated long term habit), but also because it is dif-his own perspective and that of others. Prior to his own ficult to understand how the rapid downward hedoniclong bout with depression, Styron received a visit from spiral associated with many kinds of addictions can betwo friends who were suffering from severe depression, viewed as the outcome of a rational choice. Cocainebut reports that he found their behavior baffling, since addiction, for example, seems to produce a relativelytheir depression had no reality to him in his own nonde- rapid diminution in the overall capacity for pleasurepressed state. Later, when he became depressed him- (Gardner & Lowinson, 1993). Herrnstein and Prelecself, he experienced the same empathic void with re- (1992), in contrast, argue that people become addictedspect to the people around him. because they fail to notice the small incremental nega- tive effects of the addictive substance. However their
APPLICATIONS account fails to explain why people don’t get the infor- mation from sources other than their own personal ex- perience since, as Goldstein notes, the consequences ofA major challenge confronting the decision paradigm
is the generally poor ‘‘fit’’ achieved in empirical analy- addiction are well publicized. The theoretical perspective proposed here provides ases of behavior that are guided by decision theory. In
attempts to use decision models to explain or predict somewhat different answer to this question (see, Loewenstein, forthcoming b, for a more detailed discus-such wide-ranging behaviors as job choice, migration,
contraception, criminal activity, and self-protective sion). Research on drug addiction suggests that it is not so much the pleasure of taking the drug that producesmeasures against health, home and work-place risks,
the fraction of explained variance has generally been dependency, but the pain of not taking the drug after one has become habituated to it (Gardner & Lowinson,low. Although disappointing results are often attrib-
uted to measurement error, the poor fit problem per- 1993). This pain is usually subclassified into two com- ponents: the pain of withdrawing from the drug andsists even when researchers collect their own data, and
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
286 GEORGE LOEWENSTEIN
the cravings for the drug that arise from ‘‘conditioned from sex. Clearly, the emotions associated with un- wanted pregnancy are much more powerful or at leastassociation’’ — i.e., that result from exposure to per-
sons, places, and other types of stimuli that have be- long-lasting than those associated with sexual sponta- neity and enhanced pleasure; however, and consistentcome associated with drug taking. Proposition 5 (un-
derestimation of the impact of future visceral factors) with proposition 1, the immediacy and certainty of em- barrassment and discomfort seem to overwhelm theimplies that people who have not experienced the pains
of withdrawal and craving may over- or underestimate delayed and uncertain consequences of using it or fail- ing to use it.11 Proposition 7 can, perhaps, help to ex-the aversiveness of withdrawal and craving, but will
almost surely underestimate the likely impact of these plain some of the misguided policies in this area — such as the abstinence movement — which leaves teenagersvisceral factors on their behavior. That is, people will
exaggerate their own ability to stop taking a particular unprepared for their own feelings and behavior because its proponents underestimate the influence of visceraldrug once they have started. Believing that they can
stop taking the drug at will, they are free to indulge factors on the behavior of others. their curiosity, which, according to Goldstein (1994, p.
Self-protection against sexually transmitted disease.215) is the driving force in most early drug use. Based on his own extensive and innovative researchProposition 2 can also help to explain the prevalence on the AIDS-related sexual behavior of gay men, Goldof self-binding behavior among addicts. The alcoholic (1993, p. 1994) argues that much unprotected sex oc-who takes antabuse (assuring him or herself of horrible curs in the heat of the moment but that people can’twithdrawal symptoms), the smoker who ventures off remember or predict what the heat felt like and so areinto the wilderness without cigarettes (after a final unprepared to deal with it. He believes that the poorsmoke at the departure point), and the dieter who signs memory for the ‘‘heat of the moment’’ has hamperedup for a miserable, hungry, vacation at a ‘‘fat farm’’ are researchers who ‘‘have studied only those cognitionsall imposing extreme future misery on themselves. To that are present in respondents’ minds at the time theythose who view these behaviors as the manifestation are answering the researcher’s questions (that is, ‘inof myopic time preferences, such seemingly far-sighted the cold light of day’), rather than those that are pres-behavior may seem anomalous. Proposition 2, however, ent during actual sexual encounters’’ (Gold, 1993, p. 4).suggests that such readiness to impose future pain on Based on his view that gay men forget the influence ofoneself has less to do with time preference, and more the heat of the moment (consistent with propositionto do with the unreality of future pain to the currently 6), Gold (1994) ran a study in which he compared thepain-free self. It seems unlikely that alcoholics, smok- effectiveness of a conventional informational interven-ers, or overeaters would take any of these actions at a tion intended to increase the use of condoms duringmoment when they were experiencing active craving anal intercourse (exposure to didactic posters) to a newfor the substance to which they are addicted.
Sexual Behavior 11 Immediate affect has been found to be a critical determinant of behavior in numerous analyses of decision making. For example,
As is true for addiction, volition seems to play an Grasmick, Bursik, and Kinsey (1990) conducted two surveys on lit- tering in Oklahoma City, one just before and one shortly after theambiguous and often changing role in sexual behavior. initiation of a successful anti-littering program. The survey askedAlthough we hold people accountable for their behavior people whether they littered, obtained demographics, and askedas a matter of policy, sexually motivated behavior often questions about shame (e.g., ‘‘Generally, in most situations I would
seems to fall into the ‘‘gray region’’ between pure voli- feel guilty if I were to litter the highways, streets, or a public recre- tion and pure compulsion. The following three exam- ation area’’) and also about the embarrassment the respondents
would feel if they littered. The R2 jumped from .076 to .269 whenples illustrate the applicability of the proposed theoret- shame and embarrassment variables were added to the equationical perspective to sexual behavior. predicting compliance, and the increase in these variables across the surveys mediated the change in mean compliance, strongly sug-Teenage contraception. In a recent study of teenage gesting that the effectiveness of the program was due to its success
contraceptive behavior, Loewenstein and Furstenberg in attaching an immediate negative emotion to littering. Manstead (1991) found that birth control usage was largely unre- (1995) found that age and sex (typically the two most powerful ex-
planatory variables) dropped out of regression equations predictinglated to the main variables that the decision making risk taking among drivers after controlling for affective variables.perspective would predict they should be correlated Klatzky and Loewenstein (1995) found that traditional decision mak-with — e.g., belief in birth control’s effectiveness or the ing variables (probabilities and outcome severities) explained sur-
desire to avoid pregnancy. The most important corre- prisingly little of the variance in women’s breast-self examination lates of birth control usage were embarrassment about behavior relative to subjective reports of anxiety associated with
breast cancer and self-examination.using it and perceptions that it interferes with pleasure
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
287VISCERAL FACTORS
‘‘self-justification’’ intervention. Subjects in the self- cases when people experiencing one level of a visceral factor need to make decisions for themselves when theyjustification group were sent a questionnaire which in-
structed them to recall as vividly as possible a sexual will be at a different level, rules of thumb, such as ‘‘have sex nightly, regardless of immediate desire,’’ mayencounter in which they had engaged in unprotected
anal intercourse and were asked to indicate which of provide a better guide to behavior than momentary feelings.a given a list of possible self-justifications for having
unsafe sex had been in their mind at the moment they had decided not to use a condom. They were then asked Motivation and Effort to select the self-justifications that had been in their
Another area in which the decision making perspec-mind most strongly at the time, to indicate how reason- tive falls short is its treatment of motivation and effort.able each of these seemed to them now, looking back In the decision paradigm there is no qualitative distinc-on it; and to briefly justify these responses. The men tion between choosing, say one car over another, orwere thus required to recall the thinking they had em- ‘‘deciding’’ to pick up one’s pace in the last mile of aployed in the heat of the encounter and to reflect on it marathon; both are simply decisions. Years after thein the cold light of day. The percentage of men in the decline of behaviorism, behaviorists still offer the mostthree groups who subsequently engaged in two or more coherent theoretical perspective on motivation and theacts of unprotected anal intercourse differed dramati- most sophisticated and comprehensive program of re-cally between the three groups — 42 and 41% for the search (see, e.g., Bolles, 1975).control and poster groups, but only 17% for the self-
Physical effort, and often mental effort as well, oftenjustification group. produce an aversive sensation referred to as fatigue or, at higher levels, exhaustion. Like other visceral factors,Sex lives of married couples. Recent surveys of sex-
ual behavior suggest that the sex lives of married cou- fatigue and exhaustion are directly aversive, and alter the desirability of different activities; most promi-ples tend to be even worse (in terms of frequency) than
what most people already suspected. For example, a nently, they decrease the desirability of further incre- ments of effort. Proposition 1 implies that as exhaus-recent study conducted by the National Opinion Re-
search Center (Michael, Gagnon, Laumann & Kolata, tion increases, there will be an ever-increasing gap be- tween actual and desired behavior. Anyone who has1994) found that the average frequency of intercourse
of married couples declined markedly as a function of engaged in competitive sports, or who has taught for several hours in a row can confirm this prediction; re-years of marriage. Certainly some of this drop-off re-
flects the combined effects of soured relations, dimin- gardless of the importance of performing well, and even with full knowledge that one will recover from the ex-ished attraction, etc. What is surprising, as reported
in the same study, is that many couples enjoy sex quite haustion virtually immediately after suspending the activity, sustained performance is often impossible toa lot when it actually occurs. The visceral factor per-
spective can perhaps shed some additional light on the achieve. Proposition 5 implies that people will overesti- mate their own ability to overcome the effect of fa-anomaly posed by the failure to take advantage of an
obvious opportunity for gratification. tigue — they will exaggerate the degree to which they can overcome limitations in physical conditioning, con-In the early stages of a relationship, the mere
thought of sex, or the physical proximity of the other centration, etc. through sheer willpower, and proposi- tion 6 implies that, as time passes, people increasinglypartner is sufficient to produce significant arousal. It
is easy to understand this arousal in evolutionary come to blame themselves for deficiencies in their own prior effort level because they will forget their own pastterms, and indeed research has shown that rats, cattle,
and other mammals can be sexually rejuvenated fol- exhaustion. Proposition 7 predicts that people who are observing the effort output of others will have a difficultlowing satiation by the presentation of a new partner —
the so-called ‘‘Coolidge Effect’’ (Bowles, 1974). Thus, time understanding or predicting reductions in effort output. Watching speed-skaters during the Olympics,early in a relationship one initiates sex in a visceral
state not unlike that associated with the sex act itself. for example, I found it difficult to understand why they failed to maintain their pace in the face of such over-Repeated presentation of the same sexual partner,
however, diminishes initial arousal. Proposition 5 im- whelming incentives. Many of the tactics that people use to motivate them-plies that people who are not aroused will have diffi-
culty imagining how they will feel or behave once they selves in the face of fatigue and exhaustion can be de- scribed by the observation that you can only fight vis-become aroused. It can thus explain why couples fail
to initiate sex despite ample past experience showing ceral factors with other visceral factors. Thus, a com- mon tactic for mustering willpower is to attempt tothat it will be pleasurable if they do. As in so many
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
288 GEORGE LOEWENSTEIN
imagine, as vividly as possible, the potential positive nally, Thaler and Shefrin’s (1981) ‘‘planner/doer’’ model consequences of greater effort output, or the potential adopts a principal-agent framework in which a far- negative consequences of insufficient output. When I sighted planner (the principle) attempts to reconcile lived in Boston many years ago, a friend and I would the competing demands of a series of present-oriented regularly drive to West Virginia to go canoeing, and doers (the agents). would typically drive back days later in the middle of The strength of multiple self models is that they the night. During these long drives I would remain transfer insights from a highly developed field of re- awake at the wheel by imagining myself ringing the search on interpersonal interactions to the less studied doorbell of my friend’s parents house to announce that topic of intraindividual conflict. However, the use- he had died in a car crash. The effectiveness of mental fulness of the multiple self approach is limited by im- imagery in eliciting an emotional response explains not perfections in the analogy between interpersonal and only why it is commonly used as an emotion-induction intrapersonal conflict. There is an inherent asymmetry method in research, but also may also help to explain between temporal selves that does not exist between its prominent role in decision making (c.f., Pen- different people. People often take actions that hurt nington & Hastie, 1988; Oliver, Robertson, & Mitchell, themselves materially to either reward or punish oth- 1993). Not only does imagery provide a tool for deciding ers who have helped or hurt them. In the intrapersonal between alternative courses of action but, once a reso- domain, however, people cannot take actions for the lution has been made, it may also help to stimulate the purpose of rewarding or punishing their past selves. emotional response needed to implement the decision. Another form of asymmetry arises from the fact that Multiattribute analytical evaluation seems unlikely to attempts at self control are almost always made by provide such a motivational impetus. the far sighted self against the short-sighted one, and
almost never in the opposite direction. Consistent only Self-Control with the planner-doer model, there is little camarade-
rie between successive short-sighted selves, but muchOne of the most difficult patterns of behavior to sub- more of a sense of continuity between far-sightedsume under a conventional rational choice framework, selves. For example, when people ‘‘decide’’ to sleep in,and one that has received increasing attention in the they rarely disable the alarm clock to promote theliterature, is the phenomenon of intrapersonal conflict cause of tomorrow morning’s sluggish self; however,and self-control. People sometimes report feeling as if when not actually experiencing the misery of prema-though there were two selves inside them — one more ture arousal, we might well make a policy decision topresent- and one more future-oriented — battling for place the alarm clock away from the bed every night.control of their behavior. To express the introspective
Perhaps the most significant problem with multiplesensation of intra-individual conflict, a number of peo- self models is that they are metaphorical and not de-ple have proposed different types of ‘‘multiple self’’ scriptions of what we think actually takes place in in-models that apply to intrapersonal conflict preexisting trapersonal conflict. Advocates of the multiple self ap-models that have been developed to describe strategic proach do not believe that there are little selves ininteractions between different people. people with independent motives, cognitive systems,Schelling’s multiple self model (1984), for example, and so on. Thus, it is difficult to draw connections be-constitutes a relatively straightforward application of tween multiple self models and research on brain neu-his pioneering research on commitment tactics in inter- rochemistry or physiology beyond the rather simplisticpersonal bargaining to intrapersonal conflict. In his observation that the brain is not a unitary organism.model a series of far-sighted selves who would prefer
The visceral factor perspective, and its key assump-to wake up early, eat in moderation, and desist from tion that intense visceral factors cause behavior to de-alcohol, use a variety of precommitment techniques to part from perceived self-interest, provides a better fitcontrol the behavior of their more short-sighted coun- to the stylized facts than do multiple self models. Theterparts. Elster (1985), somewhat differently, sees in- introspective feeling of multiple selves, for example,trapersonal conflict as a ‘‘collective action problem’’ in- arises from the observation that one is clearly behavingvolving the succession of one’s selves. Such a perspec- contrary to one’s own self interest. Since we are usedtive sheds special light on the phenomenon of to interpreting behavior as the outcome of a decision,unraveling. Just as one person’s cutting in line can it is natural to assume that there must be some self —cause a queue to disintegrate into a state of anarchy, other than the self that identifies one’s self-interest —the first cigarette of someone who has quit, or the first that is responsible for the deviant behavior. The factdrink of an ex-alcoholic, often usher in a resumption of
the original self-destructive pattern of behavior. Fi- that impulsive selves never promote one-another’s be-
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
289VISCERAL FACTORS
havior is not surprising if these selves are not, in fact, ity. In everyday language, the term irrationality is typi- cally applied to impulsive and self-destructive behaviorcoherent entities with consciousness and personal mo-
tives, but instead represent the motivational impact of and to actions that violate generally accepted norms about the relative importance of different goals.visceral factors. The far-sighted self, in contrast, repre-
sented by the individual’s assessment of self-interest, The theoretical perspective proposed here views irra- tionality not as an objective and well-defined phenome-is much more constant over time. The far-sighted self
can, in a sense, represent the individual’s tastes, fac- non, but as a subjective perception that occurs in the mid-range of the continuum defined by the influence oftoring out as much as possible the effect of visceral
factors. visceral factors. At low levels of visceral factors, people generally experience themselves as behaving in a ratio- nal fashion. At extremely high levels, such as the level
CONCLUDING REMARKS of sleepiness that causes one to fall asleep at the wheel, decision making is seen as arational — that is, people
The decision-making paradigm, as it has developed, don’t perceive themselves as making decisions at all. is the product of a marriage between cognitive psychol- It is in the middle region of visceral influences, when ogy and economics. From economics, decision theory people observe themselves behaving contrary to their inherited, or was socialized into, the language of prefer- own perceived self-interest, that they tend to define ences and beliefs and the religion of utility maximiza- their own behavior as irrational. Expressions such as tion that provides a unitary perspective for under- ‘‘I don’t know what got into me,’’ or ‘‘I must have been standing all behavior. From cognitive psychology, deci- crazy when I . . .’’ refer to discrepancies between be- sion theory inherited its descriptive focus, concern with havior and perceived self-interest that are produced by process, and many specific theoretical insights. Deci- the influence of visceral factor. As proposition 7 would sion theory is thus the brilliant child of equally brilliant imply, moreover, the same expressions are used to refer parents. With all its cleverness, however, decision the- to the irrational behavior of others that is difficult to ory is somewhat crippled emotionally, and thus de- comprehend as self-interested. In sum, the visceral fac- tached from the emotional and visceral richness of life. tor perspective helps to explain when and why people
Contrary to the central assumption of decision the- view their own, and others’, behavior as irrational. ory, not all behavior is volitional, and very likely most The second problem resulting from the failure to take of it is not. This is not a novel critique, but most recent account of the impact of visceral factors, is a wide- critiques along these lines have attacked from the oppo- spread skepticism toward the decision making perspec- site angle. A number of researchers have argued that tive, on the part of both the general public and of aca- most behavior is relatively ‘‘automatic’’ (Shiffrin, Du- demics in the humanities. A commonly heard complain mais & Schneider, 1981), ‘‘mindless’’ (Langer, 1989), is that decision theory fails to capture what makes peo- habitual (Ronis, Yates & Kirscht, 1989; Louis & Sutton, ple ‘‘tick,’’ or what it means to be a person (c.f., Epstein, 1991), or rule-guided (Anderson, 1987; Prelec, 1991). 1994). People who introspectively experience high con- While not disputing the importance of habitual behav- flict in their personal lives are unlikely to embrace a ior, my focus is on the opposite extreme — one that, theory of behavior that denies such conflict or that, at while perhaps less prevalent than habitual behavior, best, treats it as a matter of balancing competing rea- presents a more daunting challenge to the decision sons for behaving in different ways (Tversky & Shafir, making perspective. My argument is that much behav- 1992). The dismaying consequence of decision theory’s ior is non-volitional or only partly volitional — even in lack of general appeal is a widespread tendency for situations characterized by substantial deliberation. those in the humanities and in the general public to
The failure to incorporate the volition-undermining fall back on outmoded theoretical accounts of behavior influence of emotions and other visceral factors can be such as those proposed by Freud and his followers. The seen not only in the disappointing explanatory power of task of decision researchers, as I see it, is to try to decision models, but also in two additional significant breathe more life into decision models without losing problems faced by the decision-making perspective. the rigor and structure that are the main existing The first is the counterintuitive notion of ‘‘irrationality’’ strengths of the perspective. Incorporating the influ- that has arisen in a field which has irrationality as a ence of visceral factors, I hope, is a step in that direc- central focus. As Daniel Kahneman notes (1993), con- tion. temporary decision theorists typically define irrational-
REFERENCESity as a failure to adhere to certain axioms of choice such as transitivity or independence — a definition that Akerlof, G. A. (1991). Procrastination and obedience. American Eco-
nomic Review, 81, 1 – 19.diverges sharply from personal accounts of irrational-
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
290 GEORGE LOEWENSTEIN
Anderson, J. R. (1987). Skill acquisition: compilation of weak-method negative hedonic brain substrates activated by addicting drugs. Seminars in the Neurosciences, 5, 359 – 368.problem solutions. Psychological Review, 94, 192 – 210.
Beach, L. R. (1990). Image theory: Decision making in personal and Gawin, F. H. (1991). Cocaine addiction: Psychology and neurophysi- ology. Science, 251, 1580 – 1586.organizational contexts. Chichester: Wiley.
Becker, G., & Murphy, K. M. (1988). A theory of rational addiction. Gold, R. (1993). On the need to mind the gap: On-line versus off-line cognitions underlying sexual risk-taking. In D. Terry, C. Gal-Journal of Political Economy, 96, 675 – 700. lons, & M. McCamish (Eds.), The theory of reasoned action: ItsBiderman, A. D. (1960). Social-psychological needs and ‘involuntary’ application to AIDS preventive behavior. New York: Pergamonbehavior as illustrated by compliance in interrogation. Sociometry, Press.23, 120 – 147.
Gold, R. (1994). Why we need to rethink AIDS education for gay men.Bolles, R. C. (1975). The theory of motivation (2nd ed.). New York: Plenary address to the Second International Conference on AIDS’Harper & Row. impact: Biopsychosocial aspects of HIV infection 7 – 10 July, Brigh-
Brown, J. W. (1988). The life of the mind. Hillsdale, NJ: Erlbaum. ton, UK.
Bruner, J. (1957). Going beyond the information given. In J. Bruner Goldstein, A. (1994). Addiction: From biology to drug policy. New
et al. (Eds.), Contemporary approaches to cognition. Cambridge, York: Freeman.
MA: Harvard Univ. Press. Gould, S. J. (1992). The panda’s thumb: More reflections on natural
Chapman, C. R. (1994). Assessment of pain. In W. Nimo & G. Smith history. New York: Norton.
(Eds.), Anaesthesia. England: Blackwell. Grasmick, H. G., Bursik, R. J., & Kinsey, K. A. (1991). Shame and
Christensen, S. M., & Turner, D. R. (1993). Folk psychology and the embarrassment as deterrents to noncompliance with the law — The
philosophy of mind. Hillsdale, NJ: Erlbaum. case of an antilittering campaign. Environment and Behaviour, 23,
Christensen-Szalanski, J. J. J. (1984). Discount functions and the 233 – 251. measurement of patients’ values: Women’s decisions during child-
Gratton, A., & Wise, R. A. (1994). Drug- and behavior-associated birth, Medical Decision Making, 4, 47 – 58.
changes in dopamine-related electrochemical signals during intra- Chung, S-H., & Herrnstein, R. J. (1967). Choice and delay of rein- venous cocaine self-administration in rats. The Journal of Neuro-
forcement, Journal of the Experimental Analysis of Behavior, 10, science, 14, 4130 – 46. 67 – 74.
Herrnstein, R., & Prelec, D. (1992). Addiction. In G. Loewenstein & Cinnamon, K., & Farson, D. (1979). Cults and cons. Chicago: Nelson- J. Elster (Eds.), Choice over time. New York: Russell Sage.
Hall. Hoch, S. J., & Loewenstein, G. F. (1991). ‘‘Time-inconsistent prefer-
Damasio, A. R. (1994). Descartes’ error: Emotion, reason, and the ences and consumer self-control.’’ Journal of Consumer Research, human brain. New York: Putnam. 17, 492 – 507.
Easterbrook, J. A. (1959). The effect of emotion on cue utilization Janis, I. L. (1967). Effects of fear arousal on attitude change. In L. and the organization of behavior. Psychological Review, 66, 183 – Berkowitz (Ed.), Advances in experimental social psychology (Vol. 201. 3, pp. 167 – 224).
Elster, J. (1985). Weakness of will and the free-rider problem, Eco- Janis, I. L., & Feshbach, S. (1953). Effects of fear-arousing communi- nomics and Philosophy, 1, 231 – 265. cations. Journal of Abnormal and Social Psychology, 48, 78 – 92.
Elster, J. (1977). Ulysses and the sirens. Cambridge, England: Cam- Janis, I. L., & Leventhal, H. (1967). Human reactions to stress. In bridge Univ. Press. E. Borgatta & W. Lambert (Eds.), Handbook of personality theory
and research. Chicago: Rand McNally.Epstein, S. (1994). Integration of the cognitive and the psychody- namic unconscious. American Psychologist, 49, 709 – 24. Janis, I. L., & Mann, L. (1977). Decision making: A psychological
analysis of conflict, choice, and commitment. New York: Free Press.Fields, H. L. (1987). Pain. New York: McGraw – Hill.
Fienberg, S. E., Loftus, E. F., & Tanur, J. M. (1985). Recalling pain Kahneman, D. (1993). Presidential address to the society for judgment and decision making. St. Louis, MO.and other symptoms. Health and Society, 63, 582 – 97.
Finckenauer, J. O. (1982). Scared straight! and the panacea phenom- Katz, J., & Melzack, R. (1990). Pain ‘memories’ in phantom limbs: review and clinical observations. Pain, 43, 319 – 336.enon. Englewood Cliffs, NJ: Prentice – Hall.
Fishbein, M., & Azjen, I. (1975). Belief, attitude, intention, and behav- Kent, G. (1985). Memory of dental pain. Pain, 21, 187 – 94. ior: An introduction to theory and research. Reading, MA: Addison- Klatzky, R., & Loewenstein, G. (1995). Proximate influences and deci- Wesley. sion analyses as predictors of breast self-examination. Working pa-
per, Carnegie Mellon University, Department of Psychology.Frank, R. H. (1988). Passions within reason: The strategic role of the emotions. New York: Norton. Kosslyn, S. M., Alpert, N. M., Thompson, W. L., Maljkovic, V., Weise,
S. B., Chabris, C. F., Hamilton, S. E., Rauch, S. L., & Buonanno,Frawley, P. J. (1988). Neurobehavioral model of addiction: Addiction as a primary disease. In S. Peele (Ed.), Visions of addiction. Lexing- F. S. (1993). Visual mental imagery activates topographically orga-
nized visual cortex: PET investigations. Journal of Cognitive Neu-ton, MA: Lexington Books. roscience, 5, 263 – 87.Galanter, M. (1989). Cults: faith, healing, and coercion. New York:
Oxford Univ. Press. Kuhl, J., & Beckmann, J. (Eds.). (1994). Volition and personality. Seattle/Toronto/Bern/Göttingen: Hogrefe & Huber Publishers.Gardner, E. L. (1992). Brain reward mechanisms. In J. H. Lowinson,
P. Ruiz, R. B. Millman & J. G. Langrod (Eds.), Substance abuse: Kunreuther, H., Ginsberg, R., Miller, L., Slovic, P., Borkan, B., & Katz, N. (1978). Disaster insurance protection: Public policy les-A comprehensive textbook (2nd ed., pp. 70 – 99). Baltimore: Wil-
liams & Wilkins. sons. New York: Wiley.
Langer, E. (1989). Mindfulness. Reading, MA: Addison-Wesley.Gardner, E. L., & Lowinson, J. H. (1993). Drug craving and positive/
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
291VISCERAL FACTORS
Lewis, R. V. (1983). Scared straight — California style. Criminal Jus- market shopping behavior. Proceedings, American Psychological Association Annual Convention.tice and Behavior, 10, 209 – 226.
Nisbett, R. E., & Ross, L. (1980). Human inference: Strategies andLibet, B., Gleason, C., Wright, E., & Pearl, D. (1983). Time of con- shortcomings of social judgment. Englewood Cliffs, NJ: Prentice –scious intention to act in relation to onset of cerebral activity Hall.(readiness-potential). Brain, 106, 623 – 42.
Nisbett, R. E., & Wilson, D. D. (1977). Telling more than we canLinton, S. J. (1991). Memory for chronic pain intensity: Correlates know: Verbal reports on mental processes. Psychological Review,of accuracy. Perceptual and Motor Skills, 72, 1091 – 95. 84, 231 – 59.Loewenstein, G. (forthcoming, a). Behavioral decision theory and
Norvell, K. T., Gaston-Johansson, F., & Fridh, G. (1987). ‘‘Remem-business ethics: Skewed tradeoffs between self and other. In D. M. brance of labor pain: How valid are retrospective pain measure-Messick (Ed.), Business ethics. ments?’’ Pain, 31, 77 – 86.Loewenstein, G. (forthcoming, b). Addiction, choice, and rationality.
Olds, J., & Milner, P. (1954). Positive reinforcement produced byIn J. Elster & O. J. Skog (Eds.), Getting hooked: Rationality and electrical stimulation of septal area and other regions of rat brain.addiction. Cambridge, England: Cambridge Univ. Press. Journal of Comparative and Physiological Psychology, 47, 419 –
Loewenstein, G., & Adler, D. (1995). A bias in the prediction of tastes. 27.
Economic Journal, 105, 929 – 937. Oliver, R. L., Robertson, T. S., & Mitchell, D. J. (1993). Imaging
Loewenstein, G., & Furstenberg, F. (1991). Is teenage sexual behav- and analyzing in response to new product advertising. Journal of
ior rational? Journal of Applied Psychology, 21, 957 – 86. Advertising, 22, 35 – 50.
Loewenstein, G., & Mather, J. (1990). Dynamic processes in risk Osiatynski, W. (1992). Choroba kontroli (The disease of control).
perception. Journal of Risk and Uncertainty, 3, 155 – 75. Warszawa: Instytut Psychiatrii i Neurologii.
Lord, Lepper, & Ross (1979). Biased assimilation and attitude polar- Palm, R., Hodgson, M., Blanchard, D., & Lyons, D. (1990). Earth- ization: the effect of prior theories on subsequently considered evi- quake insurance in California. Boulder, CO: Westview Press. dence. Journal of Personality and Social Psychology, 37, 2098 –
Pennington, N., & Hastie, R. (1988). Explanation-based decision2110. making: effects of memory structure on judgment. Journal of Ex-
Louis, M. R., & Sutton, R. I. (1991). Switching cognitive gears: from perimental Psychology: Learning, Memory and Cognition, 14, 521 – habits of mind to active thinking. Human Relations, 44, 55 – 76. 33.
Mandler, G. (1964). The interruption of behavior. In D. Levine (Ed.), Pettit, P. (1991). Decision theory and folk psychology. In M. Bacha- Nebraska symposium on motivation. Lincoln, Nebraska: Univer- rach & S. Hurley (Eds.), Foundations of decision theory. Oxford: sity of Nebraska Press. Blackwell.
Manstead, A. S. R. (1995). The role of affect in behavioural decisions: Pickens, R., & Harris, W. C. (1968). ‘‘Self-administration of d-am- Integrating emotion into the theory of planned behaviour. Paper phetamine by rats.’’ Psychopharmacologia, 12, 158 – 163. presented at seminar on Affect and Decision Making. Technical Pluchik, R. (1984). A emotions: A general psychoevolutionary theory. University of Eindhoven, March 31, 1995. In K. R. Scherer & P. Ekman (Eds.), Approaches to emotion (Chap-
Mazur, J. E. (1987). An adjustment procedure for studying delayed ter 8, pp. 197 – 219). Hillsdale, NJ: Erlbaum. reinforcement, In M. L. Commons, J. E. Mazur, J. A. Nevins, & Prelec, D. (1991). Values and principles: Some limitations on tradi- H. Rachlin (Eds.), Quantitative analysis of behavior: The effect of tional economic analysis. In A. Etzioni & P. Lawrence (Eds.), Socio- delay and of intervening events on reinforcement value (Chapter economics: toward a new synthesis. New York: M. E. Sharpe. 2). Hillsdale, NJ: Erlbaum.
Pribram, K. H. (1984). Emotion: A neurobehavioral analysis, In Messiha, F. S. (1993). Fluoxetine: A spectrum of clinical applications K. R. Scherer & P. Ekman (Eds.), Approaches to emotion (Chapter
and postulates of underlying mechanisms, Neuroscience and Bio- 1, pp. 13 – 38). Hillsdale, NJ: Erlbaum. behavioral Reviews, 17(4), 385 – 396.
Rachman, S., & Arntz, A. (1991). The overprediction and underpre- Michael, R. T., Gagnon, J. H., Laumann, E. O., & Kolata, G. (1994). diction of pain. Clinical Psychology Review, 11, 339 – 55.
Sex in America: A definitive survey. Boston: Little, Brown. Rofé, Y., & Algom, D. (1985). Accuracy of remembering postdelivery Miller, G. A., Galanter, E. H., & Pribram, K. H. (1960). Plans and pain. Perceptual and Motor Skills, 60, 99 – 105.
the structure of behavior. New York: Henry Holt & Co. Ronis, D. L., Yates, J. F., & Kirscht, J. P. (1989). Attitudes, decisions, Miller, G. A., Levin, D. N., Kozak, M. J., Cook, E. W. III, McLean, and habits as determinants of repeated behavior. In A. R. Prat-
A., Jr., & Lang, P. J. (1987). Individual differences in imagery and kanis, S. J. Breckerler, & A. G. Greenwald (Eds.), Attitude, struc- the psychophysiology of emotion. Cognition and Emotion, 1, 367 – ture and function. Hillsdale, NJ: Erlbaum. 90. Ross, M. (1989). Relation of implicit theories to the construction of
Mischel, W. (1974). Processes in delay of gratification. In D. Berko- personal histories. Psychological Review, 96, 341 – 57. witz (Ed.), Advances in experimental social psychology (Vol. 7, pp. Scarry, E. (1985). The body in pain. Oxford, England: Oxford Univ. 249 – 292). Press.
Mischel, W., Shoda, Y., & Rodriguez, M. L. (1992). Delay of gratifica- Schelling, T. (1984). Self-command in practice, in policy, and in a tion in children. In G. Loewenstein & J. Elster (Eds.), Choice Over theory of rational choice. American Economic Review, 74, 1 – 11. time. New York: Russell Sage. Seeburger, F. F. (1993). Addiction and responsibility. An inquiry into
Morley, S. (1993). Vivid memory for ‘everyday’ pains. Pain, 55, 55 – the addictive mind. New York: Crossroads Press. 62. Shiffrin, R. M., Dumais, S. T., & Schneider, W. (1981). Characteris-
Morris, D. B. (1991). The culture of pain. Berkeley: University of tics of automatism. In J. Long & A. Baddeley (Eds.), Attention and California Press. performance (Vol. IX, pp. 223 – 238). Hillsdale, NJ: Erlbaum.
Siegel, S. (1979). The role of conditioning in drug tolerance and addic-Nisbett, R. E., & Kanouse, D. E. (1968). Obesity, hunger, and super-
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
292 GEORGE LOEWENSTEIN
tion. In J. D. Keehn (Ed.), Psychopathology in animals: Research Thaler, R. H., & Shefrin, H. M. (1981). An economic theory of self- control. Journal of Political Economy, 89, 392 – 406.and treatment implications. New York: Academic Press.
Simon, H. A. (1967). Motivational and emotional controls of cogni- Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for tion. Psychological Review, 74, 29 – 39. judging frequency and probability. Cognitive Psychology, 5, 207 –
Smith, A. (1892/1759). Theory of moral sentiments. London: George 32. Bell & Sons. Tversky, A., & Kahneman, D. (1991). Loss aversion in riskless choice:
Stewart, J., & Wise, R. A. (1992). Reinstatement of heroin self-admin- A reference-dependent model. Quarterly Journal of Economics, istration habits: morphine prompts and naltrexone discourages 106, 1039 – 61. renewed responding after extinction. Psychopharmacology, 108,
Tversky, A., & Shafir. (1992). Choice under conflict: The dynamics779 – 784. of deferred decision. Psychological Science, 13, 793 – 5.
Strack, F., Schwarz, N., & Gschneidinger, E. (1985). Happiness and Zajonc, R. B. (1968). Attitudinal effects of mere exposure. Journalreminiscing: The role of time perspective, affect, and mode of think-
of Personality and Social Psychology Monograph, 9, 1 – 28.ing. Journal of Personality and Social Psychology, 49, 1460 – 1469. Strotz, R. H. (1956). Myopia and inconsistency in dynamic utility Zajonc, R. B., & Markus, H. (1982). Affective and cognitive factors
in preferences. Journal of Consumer Research, 9, 123 – 31.maximization. Review of Economic Studies, 23, 165 – 80.
Received: October 4, 1995
/ a701$$2594 04-02-96 19:39:29 obha AP: OBHDP
Journal of School Psychology 1977 • Vol. 15, No. 1
A T T R A C T I V E N ESS A N D S C H O O L ACH I E V E M E N T
J O H N S A L V I A , R O B E R T A L G O Z Z I N E , a n d J O S E P H B. S H E A R E The Pennsylvania State University
Summary: Facial attractiveness has been shown to have powerfully biasing effects in hypothetical investigations of teacher attitudes. The purpose of this study was to ascertain the relationship between rated attractiveness and two measures of school performance. Attractive children received significantly higher report cards and, to some degree, higher achievement test scores than their unattractive peers. The results are discussed and implications for further research are considered.
The relationships a m o n g teachers' attitudes toward various pupil charac- teristics and pupil performance have b e e n reported in the literature. Palardy (1969) f o u n d that expectation for reading success of first-grade boys held b y teachers was significantly related to the reading performance of boys. Seaver (1973) studied the effect that a high- or low-achieving pupil had on a y o u n g e r sibling's achievement when the y o u n g e r sibling had the same or a different teacher. He c o n c l u d e d that teacher expectations did make a difference. However, because of the n o n e x p e r i m e n t a l n a t u r e of such studies, it is diffi- cult to control rival hypotheses. F o r example, the older siblings in the Seaver (1973) study could have biased their y o u n g e r siblings toward the teacher.
Recent a t t e n t i o n has been directed toward one particular s t u d e n t at- t r i b u t e - a t t r a c t i v e n e s s . Teachers have been shown to hold differential ex- pectations for attractive and u n a t t r a c t i v e children. Facial attractiveness, as inferred from photographs, has been shown to affect teacher j u d g m e n t s and decisions in h y p o t h e t i c a l situations. Kehle (Note 1) f o u n d that attractiveness was a factor in teacher ratings of fifth-grade students. Clifford and Walster (1973) f o u n d that attractive children were judged to be more intelligent by teachers who were given identical report cards with different photographs attached to them. Ross and Salvia (1975) attached photographs of attractive and unattractive children to identical, fictitious case studies of mildly handi- capped children. Experienced teachers indicated that unattractive children would have more difficulty academically and socially. F u r t h e r m o r e , these teachers favored special class p l a c e m e n t more for the u n a t t r a c t i v e than for the attractive children.
Because teachers hold differential expectancies for attractive and unattrac- tive children does n o t c o n s t i t u t e evidence that there are differences in performances of these children. Thus, a first step in ascertaining the effect of facial attractiveness in n a t u r a l settings is to learn if there is a relationship between attractiveness and pupil performance. If there is n o relationship, there is little, if any, reason to believe that the expectancies associated with facial attractiveness affect s t u d e n t performance. However, the presence of
60
A t t r a c t i v e n e s s and School A c h i e v e m e n t 61
such a r e l a t i o n s h i p does n o t i m p l y c a u s a l i t y . F i r s t , a t t r a c t i v e c h i l d r e n m a y p e r f o r m b e t t e r in s c h o o l because o f a n y n u m b e r o f factors o u t s i d e o f the schools; social class, p o o r n u t r i t i o n o r h e a l t h , and p e r s o n a l i t y variables can affect b o t h a c h i l d ' s a p p e a r a n c e and s c h o o l p e r f o r m a n c e . A s e c o n d p o s s i b i l i t y is t h a t d i f f e r e n t i a l a p p e a r a n c e c o u l d l e a d t o d i f f e r e n t i a l p a t t e r n s o f teacher- p u p i l i n t e r a c t i o n which c o u l d p r o d u c e a c h i e v e m e n t differences. A third p o s s i b i l i t y is t h a t d i f f e r e n t i a l a t t r a c t i v e n e s s could s i m p l y influence t e a c h e r appraisals o f p u p i l p e r f o r m a n c e t h r o u g h h a l o mechanisms.
The p r e s e n t s t u d y e x a m i n e d t h e r e l a t i o n s h i p b e t w e e n facial a t t r a c t i v e n e s s and p u p i l a c h i e v e m e n t in t h e school e n v i r o n m e n t . School a c h i e v e m e n t is a m u l t i d i m e n s i o n a l p h e n o m e n o n . One d i m e n s i o n is a n o r m - r e f e r e n c e d , objec- tive assessment of a t t a i n m e n t o f n a t i o n a l l y e x p e c t e d s c h o o l o u t c o m e s ; an e q u a l l y i m p o r t a n t d i m e n s i o n is the t e a c h e r ' s j u d g m e n t o f the c h i l d ' s progress t h r o u g h the local c u r r i c u l u m . Therefore, t w o measures were used in this i n v e s t i g a t i o n : scores f r o m a s t a n d a r d i z e d , group a d m i n i s t e r e d , o b j e c t i v e l y scored a c h i e v e m e n t test, and r e p o r t card grades.
METHOD
Subjects
Initial Sample. Six classes at the third-, o r f o u r t h - , and fifth-grade levels were r a n d o m l y select,',d f r o m a s c h o o l d i s t r i c t in c e n t r a l Pennsylvania. The 440 Caucasian c h i l d i e n in t h e initial sample were a p p r o x i m a t e l y e q u a l l y d i s t r i b u t e d a m o n g the 18 classrooms. A s t a n d a r d , individual, b l a c k and w h i t e p h o t o g r a p h o f each c h i l d ' s h e a d and shoulders u n d e r s t a n d a r d lighting and b a c k g r o u n d c o n d i t i o n s was t a k e n f r o m the s c h o o l d i s t r i c t ' s c u m u l a t i v e folders. These p h o t o g r a p h s were divided i n t o six same-sex same-grade groups (e.g., fifth-grade girls). Each group o f p h o t o g r a p h s was r a t e d i n d e p e n d e n t l y b y seven g r a d u a t e s t u d e n t s in special e d u c a t i o n w h o did n o t k n o w a n y o f the c h i l d r e n or the u l t i m a t e p u r p o s e o f the s t u d y . The raters were asked to divide each group i n t o five a p p r o x i m a t e l y equal s u b g r o u p s ( f r o m 12 t o 16 d e p e n d - ing on the size o f t h e p a r t i c u l a r g r o u p ) b y first selecting the least a t t r a c t i v e children, t h e n the m o s t a t t r a c t i v e c h i l d r e n , the n e x t least a t t r a c t i v e children, and the n e x t m o s t a t r r a c t i v e children.
Final Sample. The mean r a t i n g for each of the 440 c h i l d r e n was c o m - p u t e d , and t h e c h i l d r e n w i t h i n each o f t h e six grade-sex groups were t h e n ranked on t h e i r m e a n a t t r a c t i v e n e s s rating. The seven m o s t a t t r a c t i v e and the seven least a t t r a c t i v e c h i l d r e n ( a p p r o x i m a t e l y the t o p and b o t t o m 10% o f each of these groups) were selected for f u r t h e r s t u d y . I n t e r r a t e r a g r e e m e n t ( a g r e e m e n t s / d i s a g r e e m e n t s plus a g r e e m e n t s ) a m o n g the raters on the children in the final sample was .69 using the five categories o f attractiveness. The ratings for the t w o a t t r a c t i v e categories ( " m o s t a t t r a c t i v e " and " a t t r a c t i v e " ) and the t w o u n a t t r a c t i v e categories were collapsed to f o r m a three-interval c o n t i n u u m ( a t t r a c t i v e , average, u n a t t r a c t i v e ) . The r e l i a b i l i t y ( a g r e e m e n t s / agreements plus d i s a g r e e m e n t s ) for these a d j u s t e d ratings was .89, i n d i c a t i n g that it was p o s s i b l e ' t o evaluate c o n s i s t e n t l y the p h o t o g r a p h s . No child in the attractive g r o u p was ever r a t e d as u n a t t r a c t i v e ; no child in the u n a t t r a c t i v e group was ever r a t e d as attractive.
62 J o u r n a l of School Psychology
P r o c e d u r e During the spring the photographs were taken and rated. During the
s u m m e r the final report card grades (A=4, B=3, etc.) for each of the 84 children were o b t a i n e d from each child's cumulative record. The results of the Iowa Tests of Basic Skills (ITBS) (Lindquist & Hieronymus, 1956), which was r o u t i n e l y administered at the end of the school year, were also o b t a i n e d from each child's cumulative record.
R E S U L T S
ITBS. Mean grade equivalents for the groups on each subtest are presented in Table 1. A m u l t i f a c t o r (attractiveness X sex X grade X subtest) repeated measure analysis of variance was performed on the data to d e t e r m i n e the effect of attractiveness (see Table 2). A significant attractiveness X grade i n t e r a c t i o n was f o u n d (see Figure 1). A simple effects analysis for attractive- ness at each grade was t h e n performed. Only at the f o u r t h grade did attractive children receive higher scores t h a n u n a t t r a c t i v e children.
Report Cards. Means for the groups for each report card category are presented in Table 3. A m u l t i f a c t o r (attractiveness X sex X grade X category) repeated measures analysis of variance was performed to d e t e r m i n e the effect of attractiveness. As shown in Table 4, the main effect for attractiveness was significant, indicating that attractive children o b t a i n higher grades on their
Table 1 ITBS Means by Groups and Subtests (N=7 per cell)
Total Total W o r k Total
Grade and Reading Language Study Arithmetic Physical Rating Sex Vocabulary Comprehension S k i l l s S k i l l s S k i l l s Total
Grade 3 ATT
UNATT
Grade 4 ATT
UNATT
Grade 5 ATT
UNATT
Girls 3.3 4.1 3.2 3.5 3.7 3.5
Boys 2.2 2.9 2.7 3.0 3.2 2.9
Girls 4.0 4.9 4.6 4.3 4.3 4.4
Boys 2.7 3.3 3.1 2.9 3.1 3.0
Girls 5.2 6.2 5.6 5.5 5.5 5.6
Boys 3.9 4.2 4.0 3.9 4.4 4.1
Girls 4.4 4.2 4,3 4.1 4.0 4.2
Boys 3.7 4.2 3.6 4.0 3.9 3.8
Girls 4.9 5.9 5.7 5,4 5.6 5.5
Boys 5.4 5.6 5.1 5.5 5.6 5.4
Girls 4.7 5.0 4.7 4.3 4.8 4.7
Boys 4.8 5.0 5.5 5.0 4.8 4.9 i
A t t r a c t i v e n e s s a n d S c h o o l A c h i e v e m e n t
Table 2 Summary of Analysis of Variance for ITBS Data
S o u r c e
Between Subjects Attractiveness (A) Sex (B) A × B Grade (C) AX C BX C A X B X C ERROR
Within Subjects Subtest Scores(J) A X J B X J A X B X J C X J A X C X J BXCXJ A X B X C x J ERROR
ms Df f
11.250 1 2.754 50.350 1 12.325"
2.586 1 .633 124.867 2 30.567*
22.939 2 5 . 6 1 5 " 17.464 2 4 . 2 7 5 " 8.977 2 2.198 4.085 72
2.379 5 9.020* .523 5 1.984 .455 5 1.725 .310 5 1.176 .204 10 .774 .185 10 .703 .183 10 .695 .452 10 1.714 .264 360
*Significant at .05 level.
63
.5.4
5 . 2
5 . 0
4 . 8
4 . 6
4 . 4
4 . 2
4 . 0
3 . 8
3 . 6
3 . 4
3 . 2
/ s
s j j p l
J ' _ a t t r a c c l v e
. . . . . . . . . u n a t t r a c t i v e
3.0
Figure 1. Means for the Grade b y attractiveness interaction for l o w a Test o f Basic Skills Total Score.
6 4 J o u r n a l o f S c h o o l P s y c h o l o g y
Table 3 Mean Report Card b y Groups (N=7 per cell)
Grade and Social Physical Rating Sex Reading Language Spelling Arithmetic Science Studies Scholarship
Grade 3 ATT
UNATT
Grade 4 ATT
UNATT
Grade 5 ATT
UNATT
G ~ s 3.3 3.3 3.7 3.3 3.4 3.3 3.4
Boys 2.7 2.6 3.4 2.7 3.1 2.7 2.9
Gifts 3.1 3.4 4.0 3.3 3.3 3.3 3.3
Boys 2.7 2.4 3.6 2.6 2.7 2.9 2.4
Gkls 3.7 3.9 3.9 3.7 3.3 3.7 3.9
Boys 2.7 3.0 3.4 2.9 2.7 2.9 2.9
Girls 2.6 3.0 3.4 2.9 2.4 2.9 2.9
Boys 2.4 2.7 2.9 2.6 2.4 2.6 2.7
Girls 3.1 2.9 3.6 3.4 3.1 2.7 3.0
Boys 3.1 3.0 3.6 3.1 2.9 2.9 3.0
Gids 2.9 2.4 3.3 2.4 2.4 2.6 2.6
Boys 2.9 2.3 2.9 2.3 2.7 2.7 2.1
Table 4 Summary of Analysis o f Variance for Report Card Data
Source ms Df f
Between Subjects Attractiveness Sex A × B Grade A X C B X C A X B × C E R R O R
(A) (B)
(C)
Within Subjects R e p o r t Card Scores (J) A × J B × J A × B X J C X J A × C X J B X C X J A X B X C X J ERROR
21.333 1 8.497* 22.102 1 8.803*
.551 1 .219 3.216 2 1.281 3.838 2 1.529 3.709 2 1.477 1.699 2 .677 2.511 72
3.654 6 16.189' .278 6 1.231 .205 6 .909 .210 6 .929 .587 12 2 . 6 0 1 ' .217 12 .963 .116 12 .514 .137 12 .609 .225 432
*Significant at .05 level.
Attractiveness a n d School Achievement
Table 5 Summary of Analysis of Covarianee for Total Report Card Score
65
Souse ss df ms f
Attractiveness (A) Sex (B) Grade (C)
A X B A X C B X C
Error
31.77 1 31.77 5.33* 7.68 1 7.68 1.29
216.25 2 108.12 18.14" 0.04 1 0.04 .01 3.12 2 1.56 .26 1.92 2 0.96 .16
434.95 73 5.96
*Significant at .05 level.
report cards t h a n do unattractive children. To d e t e r m i n e if the effect could have been produced b y actual achievement differences, an analysis of covari- ance (Table 5) was performed using the total ITBS achievement score as the covariate and average report card grade as the d e p e n d e n t measure. Homoge- n e i t y of regression for achievement total scores o n report card totals was f o u n d to be nonsignificant. The effect of attractiveness was significant even when the effects of achievement were controlled.
DISCUSSION
In i n t e r p r e t i n g the results of this study, several things m u s t be considered. The first is that the m e t h o d of establishing facial attractiveness p r o b a b l y i n t r o d u c e d a conservative bias which operated against a n y attractiveness effects. The pictures of some average children could have been so flattering that the raters assigned t h e m to the attractive group, while the pictures of other average children could have been so u n f l a t t e r i n g that the raters assigned them to the unattractive group. If either or b o t h of these things occurred, there' would be a r e d u c t i o n in the distinctiveness of the two groups. The second factor is that individuals other t h a n the pupils' actual teachers did the rating. This is b o t h a strength and a weakness of the study. The strength is that only cues from the p h o t o g r a p h (the child's appearance) influenced the ratings; thus, a child's personality, odor, etc., had n o effect. The weakness is t h a t the children's teachers may n o t have placed all the children in the same attractiveness categories. However, b o t h weaknesses operate contrary to the hypothesis of achievement differences as a f u n c t i o n of attractiveness.
To d e m o n s t r a t e an e x p e c t a n c y effect, it would be necessary to demon- strate differences in objective as well as subjective performance indicators. Significant differences on the ITBS occurred o n l y at the f o u r t h grade. Nonsignificant differences in the anticipated direction were observed at the fifth-grade level, while nonsignificant differences in the opposite direction were observed at the third grade. Assuming that the achievement test selected by the district a d m i n i s t r a t i o n was a valid measure of what the students learned, there is n o clear support for the operation of an expectancy effect. However, the result~ do warrant further investigation.
The significant differences in the report card grades of attractive and unattractive children could have occurred for two reasons. First, the report card differences could reflect a negative stereotype. In such a case, the report
66 J o u r n a l o f S c h o o l P s y c h o l o g y
card grades d o n o t a c c u r a t e l y r e f l e c t t h e c h i l d ' s p e r f o r m a n c e b u t t h e t e a c h e r s ' biases. S e c o n d , r e p o r t c a r d grades m a y be m o r e sensitive t o a c h i e v e m e n t differences, or t h e y m a y m o r e a c c u r a t e l y r e f l e c t t h e s c h o o l ' s c u r r i c u l u m . In e i t h e r case, a t t r a c t i v e n e s s is i m p l i c a t e d . It s h o u l d be investigated as a na- t u r a l l y o c c u r r i n g biaser since it c o u l d have e i t h e r a l t e r e d t e a c h e r - p u p i l inter- a c t i o n s or altered t e a c h e r s ' p e r c e p t i o n s o f p u p i l a c h i e v e m e n t . Most s t a r t l i n g is t h e fact t h a t the r e p o r t c a r d d a t a were c o l l e c t e d in t h e last grading p e r i o d o f the year. Thus, t h e r e is a s t r o n g inference t h a t if t h e r e is a bias associated w i t h attractiveness, it is of l o n g d u r a t i o n . Previous r e p o r t s have o n l y sug- gested t h a t it was i m p o r t a n t in the f o r m a t i o n o f initial impressions (e.g., Ross & Salvia, 1975). The d a t a r e p o r t e d in t h i s ' s t u d y i n d i c a t e t h a t t h e bias m a y have been o p e r a t i v e even a f t e r 180 d a y s o f school.
The results o f this i n v e s t i g a t i o n have d e m o n s t r a t e d a r e l a t i o n s h i p b e t w e e n one s t u d e n t a t t r i b u t e (i.e., facial a t t r a c t i v e n e s s ) and one i m p o r t a n t d i m e n s i o n o f s t u d e n t p e r f o r m a n c e . As w i t h all c o r r e l a t i o n a l studies, a n u m b e r o f possible causal e x p l a n a t i o n s m a y be d e f i n e d . This s t u d y was u n d e r t a k e n as a p r e l i m i n a r y step in a t t e m p t i n g t o d e f i n e t h e r e l a t i o n s h i p b e t w e e n a t t r a c t i v e - ness ( A ) and s c h o o l p e r f o r m a n c e (B). The o b t a i n e d d i f f e r e n c e s in s c h o o l p e r f o r m a n c e (B) c o u l d have been caused b y t h e differences in perceived a t t r a c t i v e n e s s ( A ) or both c o u l d have been caused b y d i f f e r e n c e s in social class (C1), s t u d e n t p e r s o n a l i t y (C2) a n d / o r a n u m b e r o f o t h e r t h i r d factors (Cj). F u t u r e research s h o u l d a t t e m p t t o c l a r i f y w h e t h e r this r e l a t i o n s h i p is caused b y (1) bias in m e a s u r e m e n t (i.e., p u p i l e x t r o v e r s i o n , p h o t o g e n i c facial features), (2) a n t e c e d e n t c o n d i t i o n s (i.e., social class, h e a l t h ) , o r (3) t e a c h e r e x p e c t a n c i e s .
R E F E R E N C E N O T E
1. Kehle, T. J. Effect o f the student's physical attractiveness, sex, race, intelligence, and socio-economic status on teachers' expectations for student's personality and aca- demic performance. Unpublished doctoral dissertation, University of Kentucky, 1973.
R E F E R E N C E S
Clifford, M., & Walster, E. The effect o f physical attractiveness on teacher expectations. Sociology o f Education, 1973, 46, 248-258.
Lindquist, E. F., & Hieronymus, A. N. Iowa Test o f Basic Skill~ Boston: Houghton- Mifflin, 1956.
Palardy, J. What teachers believe, what children achieve. Elementary School Journal, 1969, 69, 370-374.
Ross, M., & Salvia, J. Attractiveness as a biasing factor in teacher judgments. American Journal o f Mental Deficiency, 1975, 80 (1), 96-98.
Scarer, W. B. Effects of naturally induced teacher expectancies. Journal o f Personality and Social Psychology, 1973, 28 (3), 333-342.
John Salvia Associate Professor Department of Special Education 314 Cedar Street Pennsylvania State University University Park, Pennsylvania 16802
Robert Algozzine Assistant Professor of Special Education
University of Florida Gainesville, Florida 32611
Attractiveness and School Achievement 67
Joseph Sheare Program Specialist for Emotionally Disturbed and Mildly Retarded Fairfax County Public Schools Fairfax, Virginia 22030
Received: November 3, 1975 Revision Received: April 9, 1976
Journal of Personality and Social Psychology 1993, Vol. 64, No. 3, 431-441
Copyright 1993 by the American Psychological Association, Inc. 0O22-3514/93/S3.OO
Half a Minute: Predicting Teacher Evaluations From Thin Slices of Nonverbal Behavior and Physical Attractiveness
Nalini Ambady and Robert Rosenthal
The accuracy of strangers' consensual judgments of personality based on "thin slices" of targets' nonverbal behavior were examined in relation to an ecologically valid criterion variable. In the 1st study, consensual judgments of college teachers' molar nonverbal behavior based on very brief (under 30 s) silent video clips significantly predicted global end-of-semester student evaluations of teachers. In the 2nd study, similar judgments predicted a principal's ratings of high school teachers. In the 3rd study, ratings of even thinner slices (6-s and 15-s clips) were strongly related to the criterion variables. Ratings of specific micrononverbal behaviors and ratings of teachers' physical attractiveness were not as strongly related to the criterion variable. These findings have important implications for the areas of personality judgment, impression formation, and nonverbal behavior.
The ability to form impressions of others is a critical human skill. "This remarkable capacity we possess to understand something of the character of another person, to form a concep- tion of him as a human being. . . with particular characteris- tics forming a distinct individuality is a precondition of social life" (Asch, 1946, p. 258). In the present article, we show that this capacity is even more remarkable than Asch suggested: Our consensual impressions of others, even when based on very brief observations of nonverbal behavior, can sometimes be un- expectedly accurate.
Kruglanski (1989) outlined the following definitions of the construct of accuracy in personality and social psychological research: (a) the degree of correspondence between a judgment and a criterion (the most popular definition in psychological research), (b) interpersonal consensus, and (c) a construct pos- sessing pragmatic utility. Much of the recent research on im- pression formation and personality judgment has been focused on the second definition regarding interpersonal consensus among individuals in their judgments regarding the personal- ity traits of others. This research has revealed three quite sur- prising findings regarding strangers' judgments of others. First, people can concur remarkably in some of their judgments of complete strangers, thus exhibiting high consensual accuracy (Albright, Kenny, & Malloy, 1988; Kenny, Homer, Kashy, & Chu, 1992; Paunonen, 1991). Second, these judgments can be
Part of this article is based on Nalini Ambady's doctoral dissertation conducted at Harvard University.
Preparation of this article was supported in part by the Spencer Foundation. We thank the Derek Bok Center for Teaching and Learn- ing for their cooperation and assistance, without which this research would not have been possible. We also thank Ross Buck, David Funder, Raj Marphatia, and two anonymous reviewers for their very helpful comments and suggestions.
The content of this article is solely our responsibility. Correspondence concerning this article should be addressed to Na-
lini Ambady or Robert Rosenthal, Department of Psychology, Wil- liam James Hall, Harvard University, Cambridge, Massachusetts 02138.
unexpectedly accurate in predicting targets' self-reported traits, even when they are based on very superficial interactions (Al- bright et al, 1988; Passini & Norman, 1966; Watson, 1989). Third, these impressions can be accurate, as defined both by high consensus and by the prediction of a criterion (self-rat- ings), even when there is absolutely no interaction between the targets and raters—for example, when impressions are based on 5-min videoclips of the targets (Colvin & Funder, 1991; Funder & Colvin, 1988; Watson, 1989).
In this article, we attempt to extend the research on the accu- racy of strangers' judgments in four ways. First, we assess the accuracy of consensual judgments based on exposures to ex- tremely "thin slices" of nonverbal behavior: videoclips 10 s or less in length. Second, we assess how well such consensual judgments predict a novel criterion variable. Third, we discuss how prediction of this criterion is of pragmatic utility because the criterion variable is ecologically valid and has "real life" consequences. Fourth, we examine the influence on judgmen- tal accuracy of two possible mediating variables: nonverbal be- havior and physical attractiveness.
Thin Slices of Behavior
In previous studies on the accuracy of strangers' ratings in the absence of any interaction, strangers rated targets after viewing a 5-min videotaped clip of them (Funder & Colvin, 1988). Our goal was to examine the accuracy of ratings based on much briefer noninteractive observations. Both Erving Goffman and Gordon Allport suggested that people are able to form accurate impressions of others from mere glimpses of their behavior (Allport, 1937; Goffman, 1979), and there is meta-analytic evidence demonstrating that people are unex- pectedly accurate in their consensual judgments of others from observations of thin slices of targets' expressive behavior (Am- bady & Rosenthal, 1992). Furthermore, ecological approaches to person perception and impression formation suggest that individuals communicate certain stimulus information or af- fordances that enable others to quickly form valid impressions of them (Baron & Boudreau, 1987; McArthur & Baron, 1983).
431
432 NALINI AMBADY AND ROBERT ROSENTHAL
With the goal of extending previous research, we examined the accuracy of strangers' judgments regarding personality attrib- utes from very minimal noninteractive information, that is, very brief video clips of nonverbal behavior. We also examined the effects of decreasing exposure lengths on judgmental accu- racy. Thus, we obtained ratings of video clips 10 s, 5 s, and 2 s long.
Ecologically Valid Criterion Variable
Although the findings mentioned above regarding the accu- racy of strangers' judgments are striking, the criterion variable used to assess accuracy in virtually all the studies was targets' self-ratings. An examination of criteria besides self-reports would add considerably to our knowledge concerning the accu- racy of interpersonal perception, especially because self-report data are likely to be biased (Cheek, 1982; Kagan, 1988; Wig- gins, 1973). Of particular value would be any evidence regard- ing the relationship between strangers' judgments of targets observed in a naturalistic situation and a criterion variable characterized by both pragmatic utility and ecological validity —one that is used in everyday decisions about people. Any evidence that people can make accurate judgments regarding others related to some external outcome of high ecologi- cal validity and utility would provide even more striking evi- dence for the accuracy of person perception.
To this end, we selected a criterion variable that seemed to be both ecologically valid and of considerable consequence: teacher effectiveness usually measured by any one or some com- bination of student ratings, peer ratings, and supervisor ratings. Such evaluations are important because of their practical signif- icance: They influence salary, promotion, and tenure decisions.
An additional practical objective of this study was to identify the nonverbal correlates of good teaching. Nonverbal behavior relates to the "affective" aspect of teaching (Rosenthal, 1989) and often communicates unintended yet significant informa- tion to students (Cooper & Good, 1983; Dusek, 1985; Harris & Rosenthal, 1985). Knowledge of the nonverbal correlates of ef- fective teaching would be of value not only in increasing our basic understanding of the importance of affective behavior in teaching and learning processes, but might also be of practical importance in guiding the selection and training of future teachers.
Nonverbal Behavior, Physical Attractiveness, and Judgmental Accuracy
In most of the research on interpersonal judgmental accu- racy, judgments were based on exposures to both verbal and nonverbal channels of targets' behavior. Some authors, how- ever, have suggested that the degree of consensus among and accuracy of strangers' judgments might be related to the pres- ence of observable cues, particularly nonverbal behavior and physical appearance cues (Albright et al, 1988; Kenny et al., 1992; Paunonen, 1991) and that it might be worthwhile to pro- vide strangers only with such cues and then examine the accu- racy of their judgments (Albright et al., 1988; Watson, 1989). In a meta-analysis of the accuracy of judgments based on thin slices of behavior, we found that people were quite accurate in their judgments based on expressive behavior, but we were not
able to disentangle the effects of the possible mediating vari- ables of specific nonverbal behaviors and physical attractive- ness, yet previous research has indicated that people who are nonverbally more spontaneously expressive are also judged to be more attractive, and both variables influence the accuracy of nonverbal encoding (Sabatelli & Rubin, 1986).
Clearly, nonverbal cues do influence judgments about some personality attributes (DePaulo, 1992). For example, Kenny and his associates found that certain behavioral cues such as smiling and rapid body movements coded from 20-s clips were related to observers' ratings of targets' extraversion (Kenny et al., 1992), although they did not examine the accuracy of such judgments.
Physical appearance cues, particularly physical attractive- ness, also influence judgments regarding a variety of character- istics (Berscheid & Walster, 1974; Dion, 1986), particularly those characteristics related to sociability and social compe- tence (Albright et al., 1988; Eagly, Ashmore, Makhijani, & Longo, 1991; Kenny et al., 1992). Judgments of strangers might be influenced by targets' attractiveness and might actually be accurate in assessing targets' self-ratings. One explanation for this accuracy could be that attractive targets might actually rate themselves more positively, because of the influence of self-ful- filling prophecies (Snyder, Tanke, & Berscheid, 1977).
To examine the mediating effect of nonverbal behavior and physical appearance on the accuracy of personality judgments, we controlled the information available to raters so that targets were rated solely on the basis of their nonverbal behavior. We also obtained separate judgments of the physical attractiveness of targets to examine the relationship between physical appear- ance and the accuracy of personality judgments.
Study 1: College Teachers
In the first study, we had judges rate 13 teachers on several personality dimensions on the basis of three 10-s silent video- clips of each teacher. The sample of teachers was drawn from a university setting.
The criterion was end-of-the-semester student evaluations. Although student achievement (adjusted for student ability) might be the best possible criterion of effective teaching, it is very difficult to obtain such data. Teacher effectiveness in the real world is often evaluated solely on the basis of ratings of supervisors and students. Therefore, we used end-of-the-semes- ter student ratings of teachers as a measure of teacher effective- ness. Considerable evidence supports the validity of student evaluations: Student ratings are consistent over time and across raters; correlate positively with expert, colleague, and adminis- trator ratings; are independent of extraneous characteristics or characteristics of the students themselves; correlate signifi- cantly with how much students actually learn; and, last, do not change appreciably with greater age of the student rater and reflection by the student (Abrami, dApollonia, & Cohen, 1990; Centra, 1979; Cohen, 1981; Feldman, 1989a, 1989b; Howard, Conway, & Maxwell, 1985; Kulik & Kulik, 1974; Leventhal, Perry, & Abrami, 1977; Marsh, 1984; McKeachie, 1979; Trent & Cohen, 1973). Thus, student evaluations seem to be a valid means of evaluating teacher effectiveness.
HALF A MINUTE 433
Method
Sample and Criterion
Permission was obtained to use videotapes of classes taught by 13 teachers at the Teaching Laboratory of a private university. The teachers (6 women and 7 men) were all graduate student Teaching Fellows who were videotaped while teaching sections for undergradu- ate courses at the Teaching Laboratory to receive feedback about their teaching from specialists at the center. Each section lasted approxi- mately 1 hr. The sections covered diverse areas of the curriculum: eight of the sections were in the area of the humanities and social sciences, three in the natural sciences, and two in languages. The number of students in the sections varied from 8 to over 20. The sample of teachers was selected by members of the Teaching Center from their library of videotapes. The criterion used for selection was that the sample should represent a fairly wide range in their ratings on the teaching criterion measure. The experimenters were blind to the evalu- ations of the teachers until the data analysis stage.
As mentioned earlier, we used end-of-the-semester student ratings of teachers as the criterion measure. We used the mean end-of-the-se- mester ratings of the teacher by the students in the section for which the teacher was videotaped. Ratings, compiled by the Committee on Undergraduate Education, were obtained by averaging the responses to the following items: "Rate the quality of the section overall" and "Rate section leader's performance overall." We were provided with the average of the two ratings for each teacher in our sample but not with the two separate ratings. Because some teachers were rated on a 5-point scale and others on a 7-point scale, these ratings were converted into percentages for this analysis, which varied from 41.67% to 91.67%. Only after all the data had been collected did we receive the ratings for the sample of teachers in this study.
Stimulus Material
One master videotape was derived from the 13 individual video- tapes using the following method: For each teacher, 10 s from the first 10 min of the class video that was focused on the teacher alone (without any of the students), 10 s from the middle of the class video with the teacher alone, and 10 s from the last 10 min of the class with the teacher alone were assembled and rerecorded onto one videotape. The order of the 3 clips showing the teacher alone from each of the 13 tapes was randomized in a design like a Latin square. The final stimulus tape contained 39 clips: 3 clips each of 13 teachers.
Judges and Ratings
Ratings of molar nonverbal behaviors. Nine female undergraduates were paid to rate the 39 clips of the teachers individually. Female judges were chosen because of the evidence that women are better decoders of nonverbal behavior (Hall, 1984). They were told that they would see short segments of teachers teaching a class and would be asked to rate the nonverbal behavior of the teacher in each segment on 15 dimensions on a scale running from 1 (not at all) to 9 (very). Besides these instructions, the judges received no training. Previous research has demonstrated that naive subjects are able to reliably make such ratings without training (Rosenthal, 1987; Rosenthal, Blanck, & Van- nicelli, 1984). The dimensions rated were accepting, active, anxious (this dimension was reverse scored), attentive, competent, confident, dominant, empathic, enthusiastic, honest, likable, optimistic, profes- sional, supportive, and warm. Each segment was played once with the audio turned down completely, and the judges were given enough time to complete their ratings before the next segment was shown.
Coding of molecular nonverbal behaviors. The frequencies per clip of specific molecular nonverbal behaviors were coded by two female undergraduates. For each clip the raters counted the number of head
nods, head shakes, smiles, laughs, yawns, frowns, biting of the lips, downward gazes, self-touches, fidgets by shaking the hands or legs or by fiddling with an object, emphatic gestures (pointing, clapping, etc.), and weak gestures (using hands while talking); noted the position of the hands (symmetrical or asymmetrical, folded or open), position of legs (crossed or open), and position of the torso (leaning forward or backward); and also coded whether the teacher was sitting or standing.
Ratings of physical attractiveness. As mentioned earlier, both the criterion variable (student evaluations) and the predictor variable (judges' ratings of the video clips) could be influenced by the physical appearance of the teachers. In that case any correlation between the criterion and predictor variables would be attributable to physical ap- pearance rather than expressive behavior. To evaluate the influence of physical appearance on the criterion variable, we asked two female judges to rate the physical attractiveness of each teacher on a 5-point scale on the basis of a single still video clip.
Results
Reliability of Judges' Ratings
Molar nonverbal behaviors. The reliabilities of the judges' ratings of the molar nonverbal behaviors were computed by means of intraclass correlations for all judges combined as well as for individual judges (Rosenthal, 1987). The effective reliabil- ities of the mean of nine judges' ratings ranged from .60 to .89, with a mean of .72. These reliabilities are displayed in Table 1 and were based on video clips just 30 s long.
The mean of the judges' ratings of each of the 15 molar di- mensions was computed across the 39 clips. These means were intercorrelated in a 15 X 15 correlation matrix, which was sub- jected to a principal-components analysis. The principal-com- ponents analysis before rotation indicated the presence of a single factor accounting for 71 % of the variance. On the basis of this finding, a composite variable was created, composed of 15 scales for the video channel. The anxiety variable was reverse scored as not anxious because it was negatively related to all the other molar variables. Thus, one new composite variable was
Table 1 Reliabilities of Judges' Ratings of College Teachers' Molar Nonverbal Behaviors
Molar behaviors
Accepting Active Attentive Competent Confident Dominant Empathic Enthusiastic Honest Likable (Not) anxious Optimistic Professional Supportive Warm
M Mdn Global variable
Afof9 judges
.60
.89
.60
.70
.87
.75
.71
.89
.60
.70
.67
.87
.64
.70
.63
.72
.70
.85
1 judge
.15
.47
.15
.21
.42
.25
.23
.46
.15
.21
.19
.42
.17
.21
.16
.24
.21
.36
434 NALINI AMBADY AND ROBERT ROSENTHAL
created by summing the mean rating for the video channel of 15 of the nonverbal variables with the sign of the variable taken into account. The standard deviations of the variables were similar enough, so that transformations into standard scores were not needed. The reliability of the mean of the nine judges' ratings for this new variable was .85. This reliability, an Armor's theta of .97 (an index of the internal consistency of this compos- ite; Armor, 1974; Rosenthal, 1987), and a mean intercorrelation among the 15 variables of r = .67, supported the decision to use the sum of ratings of the 15 dimensions as an additional global variable. A one-way analysis of variance (ANOVA) of this com- posite variable revealed no significant main effect for the order of the 3 clips for each teacher.
Molecular nonverbal behaviors. Some of the categories for the molecular ratings of nonverbal behavior were not observed by one or both of the two judges. These categories (biting of lips, emphatic and weak gestures, arms folded or open, and legs crossed or open) were dropped from further analysis. The effec- tive reliabilities of the remaining categories ranged from .51 to .98 (see Table 2). These ratings were subjected to a principal- components analysis, but no interpretable factors emerged. Fur- thermore, the mean intercorrelation among the molecular vari- ables (r = .38) was considerably lower than the mean intercorre- lation among the molar variables (r = .67). Behaviors were considered individually in later analyses.
Molar Nonverbal Behavior and Teacher Evaluations
Ten of the 15 molar nonverbal dimensions were significantly predictive of the criterion variable of teacher effectiveness, as can be seen from Table 3. On the whole, teachers with better evaluations were judged more favorably on the average of 15 dimensions of nonverbal behavior, r{\ 1) = .76, p < .005. Specifi- cally, teachers who were rated higher by their students were judged to be significantly more optimistic, confident, domi- nant, active, enthusiastic, likable, warm, competent, and sup- portive on the basis of their nonverbal behavior. Note also that dimensions that did not reach significance correlated substan- tially and in the predicted direction with the criterion. It is interesting that these results are consistent with other findings
Table 2 Reliabilities of Judges' Ratings oflO-s Clips of College Teachers' Specific Molecular Nonverbal Behaviors
Table 3 Correlations of Molar Nonverbal Behaviors With College Teacher Effectiveness Ratings (Student Ratings)
Variable A/of2 judges 1 judge
Sit Frown Touch head Gaze down Lean forward Fidget with hands Laugh Smile Arms symmetrical Nod head Shake head Fidget with object
M Mdn
.98
.90
.86
.85
.85
.84
.84
.80
.77
.70
.60
.51
.81
.84
.95
.81
.75
.74
.74
.73
.73
.68
.62
.55
.44
.35
.70
.73
Variable
Accepting Active Attentive Competent Confident Dominant Empathic Enthusiastic Honest Likable (Not) anxious Optimistic Professional Supportive Warm
Global variable
r
.50
.77**
.48
.56*
.82***
.79**
.45
.76**
.32
.73**
.26
.84***
.53
.55*
.67*
.76**
p<.01. ***p< .001.
based on far longer observations in suggesting that achieve- ment-oriented positive behaviors (confidence and dominance) and interpersonally oriented behaviors (warmth and support- iveness) are both related to students' ratings of teachers (Brophy & Good, 1974; Erdle, Murray, & Rushton, 1985; Rosenshine, 1971; Rosenshine & Furst, 1973; Ryans, 1960).
Molecular Nonverbal Behaviors and Teacher Evaluations
Each molecular behavior was correlated with the global over- all impression variable in an attempt to understand the specific behaviors influencing molar impressions. As can be seen from the first data column of Table 4, there was a significant negative relationship between teachers' frowning and judges' ratings of them on the overall molar variable, r(l 1) = — .61, p < .05. Simi-
Table 4 Correlations of Molecular Nonverbal Behaviors With Global Molar Behavior Rating and College Teachers' Effectiveness Ratings (Student Ratings)
Molecular nonverbal behaviors
Arms symmetrical Gaze down Frown Fidget with hands Fidget with object Nod head Shake head Laugh Lean forward Sit Smile Touch head
r with molar global rating
.21 -.11 - . 6 1 * -.39 - . 5 0
.37 -.25
.20
.02 -.21
.23 -.06
r with teacher effectiveness rating
-.18 -.34 -.46 -.70** -.56*
.38
.02
.34 -.49 -.34
.20
.12
*/><.05. * * p < . 0 1 .
HALF A MINUTE 435
larly, fidgeting (with an object or with hands) was negatively correlated with global molar impressions.
Each molecular behavior was also correlated with the crite- rion variable. Fidgeting was negatively correlated with the crite- rion variable: Teachers who fidgeted more with their hands or fiddled with an object (such as chalk or a pen) received signifi- cantly lower ratings from their students (see the second data column of Table 4). Teachers with lower ratings also tended to gaze down, frown, and lean forward more. This result seems to contradict previous research (Harris & Rosenthal, 1985; Mehrabian, 1969), in which forward lean has been associated with involvement, warmth, and so forth. However, the coding of forward leans might have been confounded with the position of the teacher, because teachers who were seated were judged as less effective. The negative correlation between leaning forward and standing, r(l 1) = -.36, ns, indicated this might have been likely, although the correlation between leaning forward and teaching effectiveness dropped only from —.49 to —.42 when sitting position was partialed out from both the variables and to -.39 when sitting position was partialed out from forward lean. Another possibility is that a teacher leaning forward might have been perceived as threatening by the students, but this needs to be explored further. Teacher effectiveness was not found to be significantly related to such behaviors as smiling, shaking the head, laughing, or touching the head, although the results sug- gest that nodding and laughing are positively related to evalua- tions of teachers (see Table 4).
Physical Attractiveness and Teacher Evaluations
The reliability of the mean of the two judges' ratings of physi- cal attractiveness was R = .80. The correlation between ratings of teachers' physical attractiveness and the criterion variable was r(l 1) = .32, ns, suggesting that students' ratings of teachers were somewhat influenced by the physical appearance of the teachers.
Partial correlation coefficients were computed to examine the relationship between physical attractiveness and the crite- rion variable, controlling for nonverbal behavior, and to exam- ine the relationship between nonverbal behavior and the crite- rion variable, controlling for physical attractiveness. Results in- dicated that controlling for physical attractiveness very slightly reduced the relationship between the global impression vari- able and the criterion, from r(l 1) = .76 to r(l 1) = .74, p < .01. Furthermore, controlling for nonverbal behavior decreased more noticeably the relationship between physical attractive- ness and the criterion, from r(l 1) = .32 to r(l 1) = .14.
The overall result of this study was quite remarkable: On the basis of observations of video clips just half a minute in length, complete strangers were able to predict quite accurately the ratings of teachers by students who had interacted with them over the course of a whole semester! Furthermore, these predic- tions retained their accuracy after we adjusted for physical ap- pearance of the teachers, indicating that the judges were pick- ing up very subtle nonverbal cues.
Study 2: High School Teachers
Because of the surprising results of the previous study and the small sample size, Study 2 was a replication of the first study
with a different population of teachers and a somewhat differ- ent criterion variable. For this study we used a sample of high school teachers.
Method
Sample and Criterion
Thirteen high school teachers (8 female and 5 male) were videotaped in the classroom at a private high school in a large city. Each class lasted 50 min. Teachers were told about the study and they volunteered to be videotaped. A camcorder was used to videotape the classes. The crite- rion measure used to assess the effectiveness of the teachers was the evaluation by the principal of the high school on a 5-point scale, in response to the request, "Rate the overall effectiveness of as a teacher." However, all the teachers received ratings of 4 or 5, possibly because the school was very selective and did not renew the contracts of teachers thought to be unsatisfactory.
Stimulus Material
One master videotape was derived from the 13 videotapes using the same method as in the previous study. The final stimulus tape was composed of 39 clips: 3 clips each of 13 teachers.
Judges and Ratings
Ratings of molar nonverbal behaviors. Eight female undergraduates were paid to rate the 39 video clips of the teachers individually on the molar dimensions described in Study 1.
Coding of molecular nonverbal behaviors. The frequencies per clip of specific molecular nonverbal behaviors described in Study 1 were coded by four undergraduates.
Ratings of physical attractiveness. As in the previous study, to evalu- ate the influence of physical appearance on the criterion variable, we asked two female judges to rate the physical attractiveness of each teacher on a 5-point scale on the basis of a single still video clip.
Results
Reliability of Judges' Ratings
Molar nonverbal behaviors. The reliabilities of the judges' ratings of the molar nonverbal behaviors were computed by using intraclass correlations (Rosenthal, 1987). The effective reliabilities of the mean of the judges' ratings for the video channel ranged from .39 to .84, with a mean reliability of .69. These reliabilities are displayed in Table 5.
As in the previous study, a composite variable was created by summing the mean rating of 14 of the nonverbal vari- ables. The variable of anxiety was dropped because of its low reliability. Again, because the standard deviations of the vari- ables were similar enough, transformations into standard scores were not needed. The reliability of the mean of the eight judges' ratings for this new variable was .80. This reliability, a mean intercorrelation among the 15 variables of .49 and an Armor's theta of .96 (an index of the internal consistency of this composite, Armor, 1974; Rosenthal, 1987), supported the deci- sion to use the sum of the same dimensions as in Study 1 as an additional global variable. A one-way ANOVA on this compos- ite variable revealed no significant main effect for the order of the three clips for each teacher.
Molecular nonverbal behaviors. The reliabilities of the mo-
436 NALINI AMBADY AND ROBERT ROSENTHAL
Table 5 Reliabilities of Judges' Ratings of High School Teachers' Molar Nonverbal Behaviors
Molar behaviors
Accepting Active Attentive Competent Confident Dominant Empathic Enthusiastic Honest Likable (Not) anxious Optimistic Professional Supportive Warm
M Mdn Global variable
judges
.64
.75
.71
.69
.59
.62
.78
.79
.60
.80
.39
.84
.49
.70
.81
.69
.79
.80
1 judge
.18
.27
.25
.24
.15
.16
.30
.31
.16
.34
.06
.40
.11
.23
.35
.24
.31
.34
Table 7 Correlations of Molar Behaviors With High School Teachers' Effectiveness Ratings (Principal's Rating)
Variable r
Accepting Active Attentive Competent Confident Dominant Empathic Enthusiastic Honest Likable Optimistic Professional Supportive Warm
Global variable* Not anxious
.64*
.41
.64*
.47
.35
.27
.53
.62*
.42
.64*
.58*
.35
.74**
.63*
.68** -.12
a The not anxious variable was dropped from the global variable be- cause of low reliability. *p<.05. **p<.0l.
lecular behaviors ranged from .31 to .94 (see Table 6). The mean reliability for the 13 coded behaviors was .74.
cance were of substantial magnitude and were related to the criterion in the predicted direction.
Molar Nonverbal Behavior and Teacher Evaluations
Seven of the 15 molar nonverbal dimensions significantly predicted the criterion variable, as can be seen from Table 7. On the whole, teachers rated more highly by the principal were judged more favorably on the average of all 14 dimensions of video nonverbal behavior, r(ll) = .68, p < .05. Specifically, teachers who were rated higher were judged to be significantly more supportive, likable, accepting, attentive, enthusiastic, warm, and optimistic on the basis of their nonverbal behavior. As in the previous study, dimensions that failed to reach signifi-
Molecular Nonverbal Behaviors and Teacher Evaluations
These variables were considered individually because a prin- cipal-components analysis revealed no interpretable factors. Moreover, the mean intercorrelation among the variables was quite low (r = .11). Correlations of the global molar variable with the individual molecular variables are shown in the first column of Table 8. Teachers who smiled more were rated signif- icantly higher on the composite global variable, r(ll) = .71, p < .05.
Teachers with higher ratings tended to be more nonverbally
Table 6 Reliability of Judges' Ratings oflO-s Clips of High School Teachers' Molecular Nonverbal Behaviors
Variable Mof2 judges 1 judge
Arms symmetrical Frown Nod head Shake head Point Sit Smile Stand Strong gesture Touch head Touch upper torso Walk Weak gesture
M Mdn
.65
.31
.59
.73
.77
.96
.86
.89
.63
.94
.84
.78
.66
.74
.77
.47
.19
.43
.58
.62
.94
.75
.80
.45
.89
.73
.64
.49
.58
.62
Table 8 Correlations of Molecular Nonverbal Behaviors With Global Molar Rating and High School Teachers' Effectiveness Rating (Principal's Rating)
Molecular nonverbal behavior
r with global molar rating
Arms symmetrical Frown Nod head Shake head Point Sit Smile Strong gesture Touch head Touch upper torso Walk Weak gesture
.26 -.06
.21
.09 -.04
.07
.71
.23 -.21
.52
.25
.27
/•with principal's
rating
.25 -.14
.25 -.11
.21 -.24
.33
.05 -.49
.53
.47
.20
HALF A MINUTE 437
active and expressive. They were more likely to walk around, touch their upper torsos, and smile (see Table 8). Less effective teachers were more likely to sit, touch their heads, and shake rather than nod their heads. Touching the head was signifi- cantly related to sitting, r(\l) = .62, p < .05, and negatively related to standing, r(l 1) = -.67, p< .05. Sitting was also signifi- cantly associated with shaking the head, HI 1) = .73, p < .01. These results suggest that teachers with higher ratings showed more nonverbal expressiveness and involvement than less effec- tive teachers.
Physical Attractiveness and Teacher Evaluations
The reliability of the mean of two judges' ratings of physical attractiveness was R = .75; the reliability of a single judge was .60. The correlation between ratings of teachers' physical attrac- tiveness and the criterion variable was ril 1) = — .18, indicating that the principal was not significantly influenced by the physi- cal appearance of teachers in making her ratings. Any correla- tion obtained between the judges' ratings and the criterion vari- able in this study could, therefore, be attributed to factors other than the appearance of the teachers.
As in the previous study, partial correlation coefficients were computed to examine the relationship between physical attrac- tiveness and the criterion variable, controlling for nonverbal behavior, and to examine the relationship between nonverbal behavior and the criterion variable, controlling for physical at- tractiveness. Controlling for physical attractiveness actually in- creased the relationship between the global impression variable and the criterion from r(l 1) = .68 to r(l 1) = .73, p < .01. Further- more, controlling for nonverbal behavior decreased the rela- tionship between physical attractiveness and the criterion from r ( l l ) = - . 1 8 t o r ( l l ) = - . 4 0 .
Thus, in this study we were able to replicate the results of the previous study with a different sample of teachers and a differ- ent criterion variable. Judges' ratings of teachers' personality based on thin slices of behavior significantly predicted the principal's evaluation of the teacher.
Study 3: "Thinning" the Slices
In the third study, we investigated whether strangers' ratings of teachers would predict the two different criterion variables from the previous studies if we thinned the slices even more. Accordingly, we had judges rate two sets of shorter video clips on the same dimensions.
Method
The original 10-s clips from the previous two studies were reduced to 5-s and 2-s clips. This was done by a research assistant unfamiliar with the ratings of the teachers or the hypotheses and results of the previous two studies. She randomly selected portions from the longer clips to comprise the four sets of the shorter clips. Thus, four new videotapes were created: two for each of the two sets of teachers, one with three 5-s clips for each teacher and the other with three 2-s clips for each teacher.
Thirty-two female undergraduates were paid to rate the four video-
tapes; eight judges rated each tape and no judge saw more than one tape. The method used was the same as in the previous two studies.
Results
The reliabilities of the eight judges and the typical single judge for each videotape are presented in Table 9. As expected, the mean reliabilities for the 2-s clips were slightly lower than the reliabilities of the 5-s clips. The means of the ratings for each dimension were summed as in the previous two studies to create a composite variable. The reliabilities of the means of the judges' ratings for the new composite variables were satisfac- tory (.90 and .73 for the college teachers, and .89 and .77 for the school teachers), as were the mean intercorrelations (rs = .85 and .61 for the 5-s and 2-s clips of the college teachers and rs = .66 and .63 for the 5-s and 2-s clips of the college teachers), and Armor's theta was also high (.95 and .94 for the college teachers and .95 for both ratings of the school teachers).
Correlations of the 15 variables as well as the composite vari- ables for the college and high school teachers' 5-s and 2-s clips are presented in Table 10. As can be seen from Table 10, the correlation of the composite based on the 5-s clips with the criterion variable was slightly higher than the correlation of the composite based on the 2-s clips for the high school teachers. For the college sample, however, there was an unexpected re- versal: The composite based on the 2-s clips correlated unex- pectedly highly with the criterion variable. However, as indi- cated below, the effect sizes within as well as between the two samples did not differ significantly.
To summarize the results of all the studies, we combined the effect sizes for ratings on the global variable (composed of the individual dimensions) using meta-analytic techniques (Ro- senthal, 1991). Table 11 displays the effect sizes of the compos- ite molar variables for each clip length for both sets of teachers. The combined overall meta-analytic effect size from all six sub- studies was .59, associated with a significant Z(Z= 5.06, p < .00001). We tested the effect sizes for clip length from each sample for heterogeneity by comparing three effect sizes for each sample, using a method for comparing correlated coeffi- cients (Meng, Rosenthal, & Rubin, 1992). For the college teacher sample the chi-square value, x2(2) = 1.52, ns, indicated that the effect sizes were homogeneous. For the schoolteacher sample the chi-square value, x2(2) = 2.06, ns, indicated that these effect sizes were also homogeneous. We also performed two contrasts on both sets of effect sizes. Contrasts testing for a linear trend using weights of 1, 0, and - 1 , for the 10-s, 5-s, and 2-s clips, respectively, were computed and were not significant for the college teacher sample (Z = .24, p = .40) or the school- teacher sample (Z = 1.14, p =. 13). Contrasts comparing the 10-s clips with the 5-s and 2-s clips using contrast weights of 2, - 1 , and - 1 , respectively, were also not significant (Z = .82, p = .21, for the college teachers and Z = 1.07, p = .14, for the school- teachers). We also compared effect sizes for the two samples for each clip length and found no significant differences between the two samples of teachers (for 10-s clips, Z = .37, p = .35; for 5-s clips, Z = .46, p = .47; and for 2-s clips, Z = 1.27, p = . 10).
These analyses show that there were no significant differ- ences in the accuracy of judgments based on video clips 10 s, 5 s, and 2 s in length. In addition, there were no significant differ-
438 NALINI AMBADY AND ROBERT ROSENTHAL
Table 9 Reliabilities of Judges' (Mean of Eight Judges' and One Judge's) Ratings of Molar Nonverbal Behaviors
Molar behavior
Accepting Active Attentive Competent Confident Dominant Empathic Enthusiastic Honest Likable (Not) anxious Optimistic Professional Supportive Warm
M Mdn Global variable
8 J s
.62
.89
.78
.75
.78
.26
.81
.92
.78
.88
.21
.90
.88
.86
.88
.75
.81
.90
College teachers
5 s
1 J
.16
.50
.30
.27
.30
.03
.35
.60
.30
.49
.03
.54
.49
.44
.49
.27
.35
.54
2 s
8 J s
.60
.83
.58
.49
.67
.58
.56
.83
.30
.71
.38
.78
.48
.67
.73
.61
.60
.73
1 J
.16
.39
.14
.11
.20
.14
.14
.39
.05
.25
.06
.30
.11
.20
.25
.16
.16
.25
5 s
8Js
.88
.87
.84
.79
.84
.77
.77
.89
.78
.85
.67
.87
.75
.88
.90
.82
.84
.89
School teachers
1 J
.49
.45
.40
.31
.40
.30
.30
.50
.30
.41
.20
.45
.27
.49
.54
.37
.40
.50
2 s
8 J s
.77
.71
.69
.46
.75
.52
.83
.79
.41
.81
.11
.86
.48
.76
.84
.65
.71
.77
1 J
.29
.25
.24
.10
.27
.13
.39
.31
.09
.35
.02
.44
.11
.29
.40
.19
.25
.30
Note. J = judge.
ences in the accuracy of judgments for the two samples of teachers.
General Discussion
These results were striking. First, we found that the ratings of complete strangers based on very thin slices of teachers' non-
Table 10 Correlations of Teacher Effectiveness Ratings With Mean Judges' Ratings of5-s and 2-s Video Clips
College teachers School
teachers
Molar behavior 5 s 2 s 5 s 2s
Accepting Active Attentive Competent Confident Dominant Empathic Enthusiastic Honest Likable Optimistic Professional Supportive Warm
Global variable3
(Not) anxious
.38
.55*
.25
.52
.58*
.40
.35
.45
.28
.34
.35
.36
.36
.50
.44
.51
.37
.69*
.58*
.57*
.62*
.38
.59*
.61*
.63*
.72*
.64*
.37
.66*
.54
.71*
.14
.44
.28
.50
.35
.48
.10
.50
.36
.43
.44
.41
.52
.40
.37
.47
.27
.31
.05
.38
.28
.17
.14
.29
.18
.26
.39
.26
.42
.33
.23
.31
.29
* The not anxious variable was dropped from the global variable be- cause of its low reliability. */><.05. **p<.0l.
verbal behavior (video clips from 2 s to 10 s long) predicted with surprising accuracy the ratings of the same teachers by people who had substantial interactions with those teachers (students and supervisors, for example). Moreover, judgments based on 30-s exposures (three 10-s clips of each teacher) were not signifi- cantly more accurate than judgments based on 6-s exposures (three 2-s clips of each teacher). We found that judges were accurate, as indicated by the high degree of consensus in their judgments, the criterion relied on by most studies on the accu- racy of personality judgment (Colvin & Funder, 1991). More remarkably, we found that judges were accurate on the basis of the high correlations between their judgments and a real-life criterion variable. These findings demonstrate the wealth of information conveyed in thin slices of behavior and the unex- pected accuracy of judgments based on these slices. Second, our results demonstrated the value of using ecologically valid criteria and everyday behavior in assessing personality and af- fective variables. Third, our results highlighted the invaluable
Table 11 Combined Effect Sizes for Clip Length and Teacher Samples
High school
Clip length College
r Combined
r Combined
10 s 5 s 2s
Combined r Combined Z
.68**
.47
.31
.50**** 2.99
.76**
.44
.71**
.66**** 4.16
.72****
.46*
.54***
.59****
3.98 2.14 2.64 5.06
* Z is the standard normal deviate correspondi: */><.05. * * p < . 0 1 . ***p<.005. ***•/?<
ling to the p value. p < . 0 0 1 .
HALF A MINUTE 439
information provided by nonverbal cues in making judgments about others.
To explain these results, we examined the role of two possible mediating variables on the accuracy of judgments from thin slices. The first variable was physical attractiveness. Contrary to the evidence regarding the biasing effects of physical appear- ance, we found that judges were not strongly influenced by the physical attractiveness of the teacher. This result could be attrib- uted to methodological differences: Studies using photographs generally report strong effects for physical attractiveness, but studies providing additional information such as expressive be- havior cues do not always find such effects (Albright et al, 1988; Bull & Rumsey, 1988). Thus, when raters actually interact with targets or when they watch video clips of targets, the effects of physical attractiveness become diluted by the other informa- tion available (Barnes & Rosenthal, 1985). For example, Zuck- erman, Miyake, & Hodgins (1991) found that unattractive voices dilute the effects of attractive faces and vice versa. An- other explanation for the relatively low influence of physical appearance cues is suggested by the work of Eagly and her colleagues, who in a meta-analysis examined the influence of physical attractiveness on beliefs about social competence, intel- lectual competence, concern for others, integrity, adjustment, and potency (Eagly et al., 1991). They concluded that physical attractiveness was strongly related to judgments of social com- petence; somewhat related to judgments of potency, adjust- ment, and intellectual competence; and unrelated to judgments regarding integrity and concern for others. The prototype of the ideal teacher used by subjects in making their judgments proba- bly involved more than just social competence. Perhaps judges rated teachers on a combination of intellectual competence (in- telligence and high task motivation), concern for others (nurtur- ance and social sensitivity), adjustment (self-esteem, mental health, and maturity), as well as social competence (sociability and likability). At an intuitive level judges may have known that "teaching can proceed quite successfully no matter how unfa- vored the teacher is by nature" (Allport, 1953, p. 875), and they may have focused on cues other than physical attractiveness in forming their impressions.
This hypothesis is supported by the significant increase in accuracy from ratings based solely on appearance (conceived of as a 0-s exposure) to ratings based on 2-s clips. This increase in accuracy is probably attributable to the presence of nonverbal movement cues, as suggested by earlier results showing that judgments based on clips with even micromomentary move- ment cues (42 ms) substantially increase the accuracy in decod- ing behavior over still frames (McLeod & Rosenthal, 1983). In addition, people who provide more spontaneous nonverbal in- formation are generally better encoders of information and are also judged to be more attractive (Sabatelli & Rubin, 1986).
Accordingly, the second mediating variable we investigated was nonverbal behavior. We found that judgmental accuracy based on thin slices of behavior was not strongly linked to spe- cific nonverbal behaviors of the targets. Judgmental accuracy was, however, strongly linked to gestalt, molar impressions based on nonverbal behavior. This result supports previous re- search showing that judgments of molar impressions, although vaguer and fuzzier, generally yield more useful information than the coding of specific behaviors. This makes intuitive sense: Although we might think that a smiling teacher should
be judged as more warm and more effective, judges interpret the smile in the overall context (Ekman, 1992; DePaulo, 1992). A smile could be interpreted as positive (genuinely warm, nur- turant, and interested), or as negative (condescending or threat- ening), depending on accompanying behaviors and the context in which the behavior occurs. But judges can identify whether a person is being warm or not, although they might not be able to isolate the specific cues driving these perceptions. An interest- ing example of the accuracy of the more fuzzy molar dimen- sions comes from the work of Holt and Luborsky (1958), who attempted to predict the success of psychiatric residents at the Menninger Clinic. They found that judges' ratings of how much they liked the resident predicted psychiatric competence better than residents' scores on a variety of projective and non- projective assessment measures.
A possible explanation for the strong influence of nonverbal behavior on impression formation from thin slices of behavior is suggested by ecological perspectives on person perception (Baron & Boudreau, 1987; McArthur & Baron, 1983). This theory posits that individuals communicate certain "affor- dances" (stimulus information) that are adaptive for perceivers to detect, and perceivers detect this information if they are attuned to detecting it. For example, perceiving characteristics associated with infancy has adaptive consequences; thus, baby- faced individuals afford or elicit certain specific reactions from people (Zebrowitz, 1990). Moreover, people typically tend to use affordance-related information when they make predic- tions regarding circumscribed contexts, as in the present study (Baron & Boudreau, 1987). Perhaps the affordances related to teaching behavior are communicated through nonverbal behav- ior, and judges are attuned to pick these up. The present results support the ecological theory of interpersonal perception in suggesting that such affordances are communicated primarily through expressive behavior and have real life consequences (Zebrowitz, 1990).
Can the results of the present study be generalized to the accuracy of impressions in everyday life? The answer is not clear. Certain settings are particularly conducive to the expres- sion and easy observation of traits; these settings include aca- demic and business settings (Kenrick, McCreath, Govern, King, & Bordin, 1990). In the present study, we were examining circumscribed, or situation-specific, accuracy (Swann, 1984): Judges were making predictions regarding a criterion variable relevant to the situation in which they were observing the tar- gets; that is, they were predicting teacher effectiveness on the basis of observations of actual teaching behavior. This setting may have contributed to increased judgmental accuracy be- cause it permitted the clear expression of relevant traits and consequently the more accurate judgment of such traits. Never- theless, these results indicate that the consensual judgments of strangers in real life can be quite accurate if they are based on observations, even very brief ones, of behaviors that are rele- vant to the criterion situation (see also Colvin & Funder, 1991).
It is also worth noting that the explicit goal of judges in this study was to make judgments about others from minimal infor- mation. This goal could have increased both the attention they paid to the stimuli as well as their motivation to be accurate (Swann, 1984). In everyday situations, people might not always be as accurate in their judgments under ordinary circum-
440 NALINI AMBADY AND ROBERT ROSENTHAL
stances. Our results, however, indicate that if people are aware that they must make a rapid judgment, these judgments are likely to be quite accurate.
These results have important implications for education. They highlight the considerable influence of very subtle affec- tive nonverbal behaviors on the teaching process and reveal that these subtle influences might be identifiable from thin slices of behavior. Teachers with higher ratings were judged more positively on affective dimensions of their personality from their nonverbal behavior, confirming the importance of teacher affect (Abrami, Leventhal, & Perry, 1982; Babad, Ber- nieri, & Rosenthal, 1989; Harris & Rosenthal, 1985; Woolfolk & Woolfolk, 1977) and suggesting that better teachers might be higher in nonverbal sending accuracy, that is, in their ability to spontaneously communicate nonverbal information (Buck, 1979). Indeed, the agreement between ratings of students who had known the teacher for a whole semester or the ratings of a supervisor who had probably known the teacher for even longer and the ratings based on thin slices suggests that some stable quality of the teacher was being communicated; impressions were consistent between different sets of raters, although the variables being judged differed. This agreement supports eco- logical perspectives on social perception indicating that raters might be tuning in to the same affordances.
These findings also suggest that teachers should be made aware of the possible impact of their nonverbal behavior and perhaps even trained in nonverbal skills (Smith, 1979; Wolf- gang, 1979; Woolfolk & Brooks, 1983). But there is little un- equivocal evidence that teaching effectiveness can be improved by training in nonverbal skills. In generalizing the results of this article, certain caveats should be kept in mind. First, judgments based on thin slices are probably most accurate for interper- sonal or affective variables (Ambady & Rosenthal, 1992). Sec- ond, such judgments should be based on actual observations of the person in the situation to which the criterion is applicable. Third, aggregate judgments should be used rather than judg- ments by single individuals.
Some limitations of this research should be kept in mind. First, multiple criteria such as peer evaluations, student evalua- tions, and supervisor ratings should be used to evaluate teacher effectiveness. In particular, student achievement and student learning should be used as a criterion in future research (Abrami et al., 1982; Feldman, 1989b), although student ratings are positively related to student achievement (Cohen, 1981; Feldman, 1989b). Second, we examined the accuracy of predic- tions regarding personality dimensions from judgments based on video clips. Future studies might examine other channels of behavior to investigate whether these results generalize to other behavioral channels such as the audio channel. Furthermore, future studies should examine whether these results generalize to other ecologically valid criterion variables.
Nevertheless, these findings are surprising and provocative. They suggest, first, that our consensual intuitive judgments might be unexpectedly accurate, and second, that we communi- cate—unwittingly—a great deal of information about our- selves. Not only do we possess the remarkable ability to form impressions of others, as Asch (1946) suggested, but, perhaps more remarkably, the impressions that we form can be quite accurate!
References
Abrami, P. C, dApollonia, S., & Cohen, P. A. (1990). Validity of student ratings of instruction: What we know and what we do not. Journal of Educational Psychology, 82, 219-231.
Abrami, P. C, Leventhal, L., & Perry, R. P. (1982). Educational seduc- tion. Review of Educational Research, 52, 446-464.
Albright, L., Kenny, D. A., & Malloy, T. E. (1988). Consensus in person- ality judgments at zero acquaintance. Journal of Personality and So- cial Psychology, 55, 387-395.
Allport, G. W. (1937). Personality: A psychological interpretation. New York: Holt.
Allport, G. W. (1953). The teaching-learning situation. Public Health Reports, 68, 875-879.
Ambady, N., & Rosenthal, R. (1992). Thin slices of behavior as predic- tors of interpersonal consequences: A meta-analysis. Psychological Bulletin, 2, 256-274.
Armor, D. J. (1974). Theta reliability and factor scaling. In H. L. Costner (Ed.), Sociological methodology, 1973-1976 (pp. 17-50). San Francisco: Jossey-Bass.
Asch, S. E. (1946). Forming impressions of personality. Journal of Ab- normal and Social Psychology, 41, 258-290.
Babad, E., Bernieri, F., & Rosenthal, R. (1989). Nonverbal communica- tion and leakage in the behavior of biased and unbiased teachers. Journal of Personality and Social Psychology, 56, 89-94.
Barnes, M., & Rosenthal, R. (1985). Interpersonal effects of experi- menter attractiveness, attire, and gender. Journal of Personality and Social Psychology, 48, 435-446.
Baron, R. M, & Boudreau, L. A. (1987). An ecological perspective on integrating personality and social psychology. Journal of Personality and Social Psychology, 53, 1222-1228.
Berscheid, E., & Walster, E. (1974). Physical attractiveness. In L. Ber- kowitz (Ed.), Advances in experimental social psychology (Vol. 7, pp. 158-216). San Diego, CA: Academic Press.
Brophy, J. E., & Good, T. L. (1974). Teacher-student relationships: Causes and consequences. New "York: Holt, Rhinehart & Wilson.
Buck, R. (1979). Measuring individual differences in the nonverbal communication of affect. Human Communication Research, 6, 47- 57.
Bull, R., & Rumsey, N. (1988). The social psychology of facial appear- ance. New York: Springer-Verlag.
Centra, J. A. (1979). Determining faculty effectiveness. San Francisco: Jossey-Bass.
Cheek, J. M. (1982). Aggregation, moderator variables, and the valid- ity of personality tests: A peer-rating study. Journal of Personality and Social Psychology, 43,1254-1269.
Cohen, P. A. (1981). Student ratings of instruction and student achieve- ment: A meta-analysis of multisection validity results. Review of Educational Research, 41, 511-517.
Colvin, C. R., & Funder, D. C. (1991). Predicting personality and behav- ior: A boundary on the acquaintanceship effect. Journal of Personal- ity and Social Psychology, 60, 884-894.
Cooper, H., & Good, T. (1983). Pygmalion grows up: Studies in the expectation communication process. New \brk: Longman.
DePaulo, B. M. (1992). Nonverbal behavior and self-presentation. Psy- chological Bulletin, 111, 203-243.
Dion, K. K. (1986). Stereotyping based on physical attractiveness: Is- sues and conceptual perspectives. In C. P. Herman, M. P. Zanna, & E. T. Higgins (Eds.), Physical appearance, stigma, and social behav- ior: The Ontario Symposium (Vol. 3, pp. 7-21). Hillsdale, NJ: Erl- baum.
Dusek, J. B. (Ed.). (1985). Teacher expectancies. Hillsdale, NJ: Erl- baum.
Eagly, A. H., Ashmore, R. D, Makhijani, M. D, & Longo, L. C. (1991). What is beautiful is good, b u t . . . . A meta-analytic review of re-
HALF A MINUTE 441
search on the physical attractiveness stereotype. Psychological Bulle- tin, 110,109-128.
Ekman, P. (1992). Telling lies (expanded ed.). New York: Norton. Erdle, S., Murray, H. G., & Rushton, J. P. (1985). Personality, classroom
behavior, and student ratings of college teaching effectiveness: A path analysis. Journal of Educational Psychology, 77, 394-407.
Feldman, K. A. (1989a). Instructional effectiveness of college teachers as judged by teachers themselves, current and former students, col- leagues, administrators, and external (neutral) observers. Research in Higher Education, 30,137-194.
Feldman, K. A. (1989b). The association between student ratings of specific instructional dimensions and student achievement: Refin- ing and extending the synthesis of data from multisection validity studies. Research in Higher Education, 30, 583-645.
Funder, D. C, & Colvin, C. R. (1988). Friends and strangers: Acquain- tanceship, agreement, and the accuracy of personality judgment. Journal of Personality and Social Psychology, 55,149-158.
Goffman, E. (1979). Gender advertisements. New York: Harper & Row. Hall, J. (1984). Nonverbal sex differences. Baltimore: Johns Hopkins
University Press. Harris, M. J., & Rosenthal, R. R. (1985). The mediation of interper-
sonal expectancy effects: 31 meta-analyses. Psychological Bulletin, 97, 363-386.
Holt, R. R., & Luborsky, L. (1958). Personality patterns of psychiatrists (Vol. 1). New York: Basic Books.
Howard, G. S., Conway, C. G, & Maxwell, S. E. (1985). Construct valid- ity of measures of college teaching effectiveness. Journal of Educa- tional Psychology, 77,187-196.
Kagan, J. (1988). The meaning of personality predicates. American Psychologist, 43, 614-620.
Kenny, D. A., Horner, C, Kashy, D. A., & Chu, L. (1992). Consensus at zero acquaintance: Replication, behavioral cues, and stability. Jour- nal of Personality and Social Psychology, 62, 88-97.
Kenrick, D. T, McCreath, H. E., Govern, J., King, R., & Bordin, J. (1990). Person-environment intersections: Everyday settings and common trait dimensions. Journal of Personality and Social Psychol- ogy, 58, 685-698.
Kruglanski, A. W (1989). The psychology of being "right": The prob- lem of accuracy in social perception and cognition. Psychological Bulletin, 106, 395-409.
Kulik, J. A., & Kulik, C. C. (1974). Student ratings of instruction. Teaching of Psychology, 1, 51-57.
Leventhal, L., Perry, R. P., & Abrami, P. C. (1977). Effects of lecturer quality and student perception of lecturer's experience on teacher ratings and student achievement. Journal of Educational Psychology, 69, 360-374.
Marsh, H. W (1984). Student's evaluation of university teaching: Di- mensionality, reliability, validity, potential biases and utility. Journal of Educational Psychology, 76, 707-754.
McArthur, L. Z., & Baron, R. M. (1983). Toward an ecological theory of social perception. Psychological Review, 90, 215-238.
McKeachie, W J. (1979). Student ratings of faculty: A reprise. Aca- deme, 65, 384-397.
McLeod, P. L., & Rosenthal, R. (1983). Micromomentary movement and the decoding of face and body cues. Journal of Nonverbal Behav- ior, 8, 83-90.
Mehrabian, A. (1969). Significance of posture and position in the com- munication of attitude and status relationships. Psychological Bulle- tin, 71, 359-372.
Meng, X., Rosenthal, R., & Rubin, D. R. (1992). Comparing correlated correlation coefficients. Psychological Bulletin, 111, 172-175.
Passini, F. T, & Norman, W T. (1966). A universal conception of per- sonality structure? Journal of Personality and Social Psychology, 4, 44-49.
Paunonen, S. V (1991). On the accuracy of ratings of personality by strangers. Journal of Personality and Social Psychology, 61,41X-411.
Rosenshine, B. (1971). Teaching behaviors and student achievement. London: National Foundation for Educational Research.
Rosenshine, B., & Furst, N. (1973). The use of direct observation to study teaching. In R. Travers (Ed.), Second handbook of research on teaching. Chicago: Rand McNally.
Rosenthal, R. (1987). Judgment studies. New York: Cambridge Univer- sity Press.
Rosenthal, R. (1989, August). Experimenter expectancy, covertcommu- nication, and meta-analytic methods. Donald T. Campbell award ad- dress presented at the 97th Annual Convention of the American Psychological Association, New Orleans, LA.
Rosenthal, R. (1991). Meta-analytic procedures for social research (Rev. ed.). Newbury Park, CA: Sage.
Rosenthal, R., Blanck, P. D, & Vannicelli, M. (1984). Speaking to and about patients: Predicting therapists' tone of voice. Journal of Con- sulting and Clinical Psychology, 52, 679-686.
Ryans, D. G. (1960). Characteristics of teachers. Washington, DC: Amer- ican Council on Education.
Sabatelli, R. M., & Rubin, M. (1986). Nonverbal expressiveness and physical attractiveness as mediators of interpersonal perceptions. Journal of Nonverbal Behavior, 10,120-133.
Snyder, M., Tanke, E., & Berscheid, E. (1977). Social perception and interpersonal behavior: On the self-fulfilling nature of social stereo- types. Journal of Personality and Social Psychology, 44, 656-666.
Smith, H. A. (1979). Nonverbal communication in teaching. Review of Educational Research, 49, 631-672.
Swann, W B., Jr. (1984). Quest for accuracy in person perception: A matter of pragmatics. Psychological Review, 91, 457-477.
Trent, J. W, & Cohen, A. M. (1973). Research on teaching in higher education. In R. Travers (Ed.), Second handbook of research on teach- ing (pp. 997-1071). Chicago: Rand McNally.
Watson, D. (1989). Strangers' ratings of five robust personality factors: Evidence of a surprising convergence with self-report. Journal of Personality and Social Psychology, 57, 120-128.
Wiggins, J. S. (1973). Personality and prediction: Principles of personal- ity assessment. Reading, MA: Addison-Wesley.
Wolfgang, A. (1979). The teacher and nonverbal behavior in the multi- cultural classroom. In A. Wolfgang (Ed.), Nonverbal behavior: Appli- cations and cultural implications (pp. 159-174). San Diego, CA: Aca- demic Press.
Woolfolk, A. E., & Brooks, D. M. (1983). Nonverbal communication in teaching. In E. Gordon (Ed.), Review ofresearch in education (Vol. 10, pp. 103-150). Washington, DC: American Educational Research Association.
Woolfolk, R. L., & Woolfolk, A. E. (1977). Effects of teacher verbal and nonverbal behavior on student perceptions and attitudes. American Educational Research Journal, 11, 297-303.
Zebrowitz, L. A. (1990). Social perception. Pacific Grove, CA: Brooks/ Cole.
Zuckerman, M., Miyake, K., & Hodgins, H. S. (1991). Cross-channel effects of vocal and physical attractiveness and their implications for interpersonal perception. Journal of Personality and Social Psychol- ogy, 60, 545-554.
Received April 6,1992 Revision received August 5,1992
Accepted September 26,1992 •
Attention versus Distraction: The Interactive Effect of Program Involvement and Attentional Devices on Commercial Processing Author(s): Kenneth R. Lord and Robert E. Burnkrant Source: Journal of Advertising, Vol. 22, No. 1 (Mar., 1993), pp. 47-60 Published by: Taylor & Francis, Ltd. Stable URL: http://www.jstor.org/stable/4188870 Accessed: 25-03-2018 15:15 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Taylor & Francis, Ltd. is collaborating with JSTOR to digitize, preserve and extend access to Journal of Advertising
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
Attention Versus Distraction:
The Interactive Effect of Program Involvement
And Attentional Devices on Commercial Processing
Kenneth R. Lord and Robert E. Burnkrant
Viewer processing of television commercials is influenced by an affective triad of viewer involvement; that is, involvement in a particular program interacts with both the viewer's involvement in the advertising message and with the ad's inherent attention-engaging capacity. Two experiments were conducted to test hypotheses relevant to this theory; the results support the assertion that both program involvement and dramatic attentional devices used in television advertisements, based on their interaction with one another and with viewers'processing motivation, have the capacity to enhance or hinder the generation of viewer thoughts relevant to ad messages. Building on prior studies, this research reveals some of the cognitive processes underlying program involvement as well as demonstrating the interactive nature of that construct in terms of message and audience characteristics.
Introduction
"Super Bowl's Sponsors Watch Gulf Warily"
"Agencies Scramble to Prove Ads Aren't Hurt by TV War Coverage'
"Slip-Ups Can Put Ads in Right Place at the Wrong Time"
The above headlines, culled from The Wall Street Journal, illustrate an increasing sensitivity to the potential of program or editorial context to influence the way in which consumers process advertisements. Like- wise, recent scholarly research has begun to invite inquiry into the role of context effects; for example, Anand and Sternthal's (1990) call for research investigating "how ease of processing is affected by ... envi- ronmental factors such as the nature of the program in which advertis- ing is presented" (p. 352).
While viewers may be accustomed to being whip-sawed on a roller- coaster ride over life-and-death perils and advertising jingles, the de- sirability and effects of such a practice are unclear and subject to some debate. Understanding whether program involvement acts (either by itself or in combination with consumer or advertisement characteristics) to intensify or inhibit ad processing is becoming increasingly important to advertisers in an era in which clutter has increased, audiences have splintered, and "zapping" has become a household word.
This paper examines how a television viewer's involvement in pro- gram content affects ad-related thoughts, when an advertising message comes onto the screen. It also attempts to capture the way in which program involvement interacts with, or is moderated by, a viewer's motivation to process the ad and by the message's own attention-engag- ing capacity. In keeping with this focus, the results of two experiments using ads characterized by high and low involvement serve to detail the impact of high and low program involvement levels; the presence (or absence) of common devices employed by advertisers to attract viewer attention is also investigated in this context.
Kenneth R. Lord (Ph.D., The Ohio State University) is assistant professor of marketing, State University of New York at Buffalo.
Robert E. Burnkrant (Ph.D., The University of Illinois) is professor of marketing at The Ohio State University.
This study was supported in part by a Graduate Student Alumni Research Award from The Ohio State University and a Summer Research Fellowship from the University at Buffalo School of Management. The authors wish to thank Wayne Talarzyk, Peter Dickson, Paul Miriard, Paul Sauer and three anonymous Journal of Advertising reviewers for their valued suggestions. Appreciation is also extended to Robert Owen, Sonia Tewari and Ben Liu for assistance in stimulus development, data collection, and analysis.
Journal of Advertising, Volume XXII, Number 1
March, 1993
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
48 Journal of Advertising
Program Involvement Literature
Empirical findings regarding the effect of viewer involvement in television programming are varied. It has been argued by some that a high level of program involvement should inhibit viewer processing of com- mercial messages (Kennedy 1971; Bryant and Comisky 1978; Soldow and Principe 1981; Thorson et al. 1985). This conclusion is based primarily on a finding, generally consistent among the cited studies, that viewers demonstrated better recall for and learn- ing of commercial message content when program involvement was low than when it was high. For example, Bryant and Comisky (1978) found recall highest for a Hamm's beer commercial when it inter- rupted a minimally involving non-program control stimulus (a moving wave pattern), with recall de- creasing somewhat for moderately involving positions in "Banacek," an action-adventure show, and falling substantially at the point of climax for the same pro- gram. An alternative view (Krugman 1983) is that commercials increase in their persuasive impact as the level of interest in the program goes up. Krugman's conclusion stems from a finding of less positive viewer attitudes toward a commercial sponsor when adver- tisements appeared in a "natural break" in a program (operationalized as Barbara Walters and variety seg- ments, assumed to be low in involvement) than when they were aired in an "interrupted context" ("G.E. Theater' and documentaries, assumed to be high in involvement).
On the basis of response times observed prior to and during commercial breaks, Lord and Burnkrant (1988a) concluded that certain program content has the capacity to induce viewers to commit a large pro- portion of available attentional resources to its pro- cessing, thereby minimizing the efficiency with which they can encode and store information conveyed by a commercial. This result is consistent with, and ap- pears to account for, the recall results obtained in earlier studies. In developing a conceptual model to explain "program elaboration" effects, Lord and Burnkrant (1988b) postulated that the presence of an attention-engaging device would moderate the ex- pected impact of high program involvement.
More recently, Murry, Lastovicka and Singh (1992) demonstrated that program involvement had an effect on attitude toward the ad that was independent of program-elicited feelings and liking for the program. While they did not identify the direction of this effect or discuss its implications, its presence highlights the importance of obtaining an enhanced understanding
of program involvement effects to complement the growing body of literature which addresses the im- pact of program-induced affect on commercial pro- cessing (Burnkant, Unnava and Lord 1992; Gardner 1985; Goldberg and Gorn 1987; Pavelchak, Antil and Munch 1988; Schumann 1986).
Cognitive Arousal and Attentional Focus
Central to an understanding of program context effects on the processing of television commercials is a recognition of the capacity constraints of short-term memory (STM). While the precise size of working memory (with respect to the number of pieces of in- formation which it can accommodate at a single time) is unknown and potentially variable, it is clearly finite and quite limited (Miller 1956; Simon 1974). In view of this capacity limitation, one must selectively isolate, from the enormous set of stimuli present in the envi- ronment, those few which are to become the subject of STM operations. That selection process may be vol- untary (volitional, purposeful) or involuntary (imposed by some element of surprise or contrast with compet- ing stimuli).
A television viewer purposely commits some of his/ her limited attentional capacity to the processing of the program. The amount of capacity thus activated may vary substantially across programs or program segments, from nearly total cognitive passivity (mini- mal attentional resources employed in processing the program - for example, a parent entrapped with his or her child in front of a television set emitting an episode of "Teenage Mutant Ninja Turtles') to those inducing extensive rehearsal (mental repetition of information) and elaboration (integration between stimulus information and knowledge stored in memory). Which cognitive environment - the more highly charged or the more passive - best facilitates the processing of the commercial? The answer may depend on the relative balance of three potentially countervailing forces - the program context, the ad- vertisement itself, and the recipient's motivation to process the ad.
An audience immersed in a highly involving televi- sion program will potentially have a higher level of attentional resources activated at the outset of the advertisement than one which begins its exposure in a setting engendering minimal cognitive response (low involvement programming). Since some degree of cognitive arousal is required for active processing of stimuli to occur, high program involvement should,
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
March 1993 49
to some extent, enhance the viewer's opportunity to process a commercial (MacInnis, Moorman and Jaworski 1991) by removing the obstacle of mindless passivity. However, as the simple processes of at- tending to and encoding program content give way to the more complex tasks of program rehearsal and elaboration, the amount of attentional capacity available for the processing of subsequent television action (such as an ad) is reduced. The apparent dis- advantage of high program involvement, then, is that while it effectively activates attentional resources, it may interfere with viewers' ability to process an ad (McInnis, Morrman and Jaworski 1991) by focusing those resources on a contextual stimulus (the program) at the expense of the desired shift in attention toward the new message. W7hile a message presented in a passive environment may benefit from injecting itself as the focus of attention without having to compete with elaboration centered on another stimulus, the short timne span for most ads may be insufficient for activating the attentional resources requisite for elaborative processing of the message. Thus, the en- trapped parent described earlier may find himself or herself so cognitively understimulated by "The Turtles" as to allocate minimal attention and virtually no message-relevant thought to an accompanying ad. The same adult may be aroused to a more active cognitive state by the more engaging (but still not grippingly thought-provoking) plot and characters of a situation comedy targeted to adults. But, when the viewer is anticipating the next play in a tied game or is outraged by a story on the news, the individual may be too consumed with these thoughts to attend to the processing of an ad.
Advertisements vary with respect to their ability to engage the attentional resources and focus of their audience. A strategy sometimes employed by adver- tisers is to begin an ad with a dramatic, novel, or otherwise unexpected visual and/or audio stimulus which imposes itself upon the senses of its audience with such intensity that it commands immediate at- tention, at least for an instant, even in the absence of any intrinsic or extrinsic sources of personal relevance (Celsi and Olson 1988). Advertisers obviously hope such a ploy will result in increased processing of their message, either by activating previously inert attentional resources or by shifting the focus of re- sources already activated. This is consistent with Berlyne's (1960) observation that in situations of re- sponse conflict the more arousing stimulus tends to seize control of behavior. Alternatively, Kahneman (1973) notes that "novel and surprising stimuli," while
attracting attention, also require greater effort to pro- cess than do more familiar stimuli. An attention-en- gaging device may therefore be successful in divert- ing a viewer's attention away from the program but still fail to reorient it toward the commercial. In- stead, the effort involved in processing the device itself could serve to further distract the viewer from the processing of the advertisement's message con- tent.
A third determinant of cognitive arousal and attentional focus is the consumer's motivation to process the advertisement (MacInnis, Moorman and Jaworski 1991). One may be motivated to process an ad in some depth because of an inherent personal interest in its topic (intrinsic source of personal rel- evance; e.g., high product involvement). Alternatively, processing motivation may be heightened upon a par- ticular occasion because of some external inducement (extrinsic sources of personal relevance) present in a situational context such as a research setting (e.g., cover story or instructions that create an expectation that questions will be asked about the ad). Whatever its source (intrinsic or extrinsic), this involvement may be expected to heighten and provide direction to the cognitive processing of the ad (Batra and Ray 1983). Thus, a television viewer for whom a commer- cial has personal relevance (high intrinsic involve- ment) will voluntarily attend to the ad (voluntary selection). A viewer who is low in intrinsic involve- ment, on the other hand, has little inherent process- ing motivation but may experience an increase in motivation as a consequence of some extrinsic in- ducement (Celsi and Olson 1988; Peter and Olson 1990).
This research focuses on the effects of both program and ad involvement (either intrinsic involvement based on enduring interest in and relevance of the message topic or extrinsic experimentally-induced involvement) on the cognitive responses of the audi- ence. The research also examines the moderating role of attention-engaging devices and their effects on cognitive responses.
Hypotheses
The foregoing discussion implies the presence of different programming contexts and advertising ap- proaches, both varying in their capacity to capture the attentional resources and focus of their viewing audience, and reaching viewers whose own motivation to process the commercials may be relatively high or low. How do these three factors interact to facilitate
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
50 Journal of Advertising
or hinder the efficient processing of commercial mes- sages?
Given the focus of this research on cognitive pro-
cesses underlying persuasion, rather than on the out- come variables more commonly examined in extant
program involvement literature (i.e., recall and atti- tude), hypotheses are expressed in terms of predicted levels of ad-relevant thought. This is consistent with the demonstrated role of cognitive responses in medi-
ating persuasion (cf. Greenwald 1968; Wright 1973),
and extends the contribution of program involvement literature by revealing the impact of the predicted
interactions at the process level. This choice of de-
pendent variable (cognitive responses) admittedly constrains the present investigation's capacity to de-
ted processing effects primarily to the cognitive arena. Cognitive response measures can detect traces of such simplifying heuristics as peripheral route processing (Petty and Cacioppo 1986), categorical inferences (Fiske and Pavelchak 1986), schematic processing (Chaiken, Liberman and Eagly 1989), and affective responses (Batra and Ray 1986). However, designed as they are to tap the output of cognitive processing, they clearly perform best as indicators of processing based on the purposeful allocation of attentional ca- pacity to the relevant stimulus. While some attitude/ preference effects may occur in the absence of cogni- tive response (Nord and Peter 1980), this research assumes that most advertising messages stand to benefit from at least some cognitive processing on the part of their intended audience, thus rendering an understanding of the interaction of message and context effects on such processing a valid focus for meaningful advertising research.
A viewer who is low in intrinsic ad involvement has little inherent motivation to pay attention to a com- mercial message. In the absence of program involve- ment, an attention-engaging device at the outset of the advertisement, or some external inducement (causing extrinsic ad involvement), there will be little to break the inertia of cognitive passivity, and ad elaboration is unlikely. Placing the viewer into a state of higher cognitive arousal - either through exposure to highly involving programming prior to the com- mercial break or by means of an attentional device at the outset of the advertisement - may render ad elaboration more likely than in an environment characterized by more complete passivity, simply be- cause of the higher activation of attentional resources, as in the examples cited earlier. However, high pro- gram involvement coupled with an attention-engaging device may induce distraction which, in the absence
of personal motivation to process the message (low ad
involvement), may preclude meaningful message elaboration. For example, a football fan contemplat-
ing the next play in a fourth-and-goal situation, if interrupted by a low-involvement commercial which opens with an attention-engaging gag, has at least two things competing for his attention which may distract him or her from attending to the ensuing message details - the ongoing thinking about the
fourth-and-goal play and the humorous lead-in of the commercial. These considerations are reflected in the
following research hypotheses: H ,: When ad and program involvement are
both low, viewers exposed to an attentional device at the outset of a com- mercial will generate more ad-relevant thought than those exposed to the same advertisement in the absence of the at- tention-engaging mechanism.
H2: When ad involvement is low and program involvement is high, viewers exposed to
an attentional device at the outset of a commercial will generate less ad-relevant
thought than those exposed to the same
advertisement in the absence of the at-
tention-engaging mechanism. A viewer who is highly involved in a commercial
presumably possesses sufficient processing motiva- tion to voluntarily commit attentional resources to
that task, even in an otherwise cognitively passive environment (low program involvement). Indeed, the individual may be expected to more fully elaborate upon the ad in the absence of an attention-engaging device than in its presence, because the inclusion of an attentional stimulus may cause thoughts which would otherwise focus on the topic of the ad and its personal relevance to be briefly diverted to the device itself. For example, a health-conscious viewer who is not otherwise distracted by ongoing thinking about a program could have more opportunity for acting upon his or her intrinsic motivation to process an ad about a new nutritional marvel if the ad moves immedi-
ately to the information of interest than if it starts with a loud noise or humorous or sentimental tale which could become the focus of attention and elabo-
ration (briefly delaying processing of essential message content).
H3: When ad involvement is high and pro- gram involvement is low, viewers exposed to an attentional device at the outset of a
commercial will generate less ad-relevant thought than those exposed to the same
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
March 1993 51
Figure A
Attentional Device Effects
Program Involvement
Ad Involvement Low High
High Distraction Redirect Attention To Ad
Low Raise Cognitive Distraction Arousal
advertisement in the absence of the at-
tention-engaging mechanism. As in the low ad-involvement situation, high pro-
gram involvement poses some competition for the ad- vertising message with respect to attentional focus, given the likelihood of ongoing program-relevant
thought. But when ad involvement is high, each mil- lisecond of ongoing program elaboration which occurs before the viewer detects the message's relevance to
his/her intrinsic interests (or extrinsic processing goals) does so at the expense of ad-relevant thought. While high ad involvement may ultimately be ex- pected to induce the viewer to self select the adver-
tising message, intense ongoing elaboration of pro- gram content may create a lapse between the initial presentation on the screen of the commercial message and the viewer's observing that it addresses a topic of personal relevance. Self-selection cannot occur until the viewer is consciously aware of the presence and nature of the alternative stimulus competing for at- tention. In such a scenario an attention-engaging device may facilitate commercial elaboration by in- terrupting program-relevant thought sufficiently to expedite an awareness of the presence of the highly
involving ad. And despite the distraction potential of the device's combination with high program involve-
ment, the viewer's processing motivation should en- able him or her to move immediately past the dis-
traction and into more meaningful elaboration of the ad itself. Thus, the health-conscious viewer described
above will benefit from the inclusion of an engaging opening if it successfully breaks the stranglehold the program has on his or her thoughts. Accordingly, an attentional device is expected to facilitate ad process- ing when a highly involving ad appears at a moment of high program involvement:
H4: When ad and program involvement are both high, viewers exposed to an attentional device at the outset of a com-
mercial will generate more ad-relevant thought than those exposed to the same advertisement in the absence of the at-
tention-engaging mechanism.
The hypothesized role of attentional devices, under the various program and ad involvement conditions, is summarized in Figure A.
Experiment 1
Research Design
A 2-by-2 factorial design was employed, with ma- nipulations of program involvement (at high and low levels) and attention-engaging devices (present or absent). The design was replicated across two com- mercials, one for a low involvement product and one for an issue of high personal relevance to the student sample employed in the study.
To establish relatively high and low involvement positions within a television program, subjects in a pretest provided concurrent task response latency measures as they viewed the test program. This pro- cedure involved recording the time it took subjects to push a button in response to an audio or visual stimulus, with longer response times indicating the increased use of attentional capacity for the primary task (viewing the program). This measurement ap- proach has seen extensive use and validation in edu- cational psychology research (Britton et al. 1982), and recent consideration, utilization and validation in a marketing context (Burnkrant and Sawyer 1983; Moore, Hausknecht and Thamodaran 1986; Lord and
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
52 Journal of Advertising
Burnkrant 1988a; Lord, Burnkrant and Owen 1989).
Forty-nine undergraduate students viewed an epi- sode of 'Alfred Hitchcock Presents," with computer-
generated tones dubbed onto the audio track at ran-
dom three- to nine-second intervals. Visually isolated from one another, the subjects pressed a hand-held button each time a tone was heard, and a computerized timer recorded their response time in milliseconds.
Longer response times reflected increased program
involvement. The mean response time associated with the position selected as the low program involvement
manipulation was .475 seconds, while that for the high involvement position was .816 seconds, a signifi-
cant difference (t = 3.82, p < .01). The low program involvement position appeared seven minutes into the program. Some casual conversation had taken place at an industrial site, where the workers were about to break for lunch. No thought-provoking or
exciting action had occurred. The high involvement position occurred six minutes before the end of the
program, when viewers witnessed a series of events which appeared to be leading to the imminent poi- soning of the protagonist. (A consistent pattern of longer response time at the end of the program could reflect viewer tedium rather than elevated program involvement. This is clearly not the case in this in- stance, however, since faster response times both
preceded and followed the peak of dramatic intensity which was selected for the high program involvement
condition.) The commercials employed in the experiment were:
(1) an advertisement for Mobil 1, a premium synthetic motor oil (low intrinsic ad involvement); and (2) a message urging viewers not to allow their friends who drink to drive under the influence of alcohol (high intrinsic ad involvement). Both were profes- sionally produced commercials and neither had been aired recently in the viewing area from which sub- jects were drawn.
The attention-engaging mechanism employed with the Mobil 1 ad in the condition calling for an attentional device was a series of six audio beeps leading into the message. Though the beeps were created separately and appended to the existing ad, this is not an unreasonable or uncommon approach to attention solicitation in advertising. For example, a recent ad for Braun electric shavers employs a com- parable device early in the message. As a pretest of this device, twenty-five undergraduate business stu- dents were exposed to the Mobil 1 ad, removed from the program context. Eleven subjects saw the ad pre- ceded by the tones, and the remainder saw the ad
without them. Those in the device-present condition perceived the ad to be realistic (mean 5.64 on a seven- point scale); all recalled the tones. Additionally, a substantive difference emerged in the number of commercial elements recalled in the two conditions (means of 9.91 with the beeps and 7.93 in their ab- sence), and this too is indicative of the device's at- tention-engaging capacity. It was thus concluded that the tones served as an effective attention-engaging device.
For the drinking and driving advertisement, the introductory attentional stimulus was a stark visual image of a living human hand shaking a skeletal hand, and the sound of screeching brakes and an automobile crash. The following words were then flashed on the screen: "Drinking and driving can kill a friendship." Direct discussion of the crash scene with several individuals who viewed it in the early stages of stimulus development, and the observation of the responses of a few individuals in a viewing setting, revealed that this device engaged viewers' attention and elicited comments relative to its pow- erful imagery. To keep time and amount of information constant, the scene was incorporated into all versions of the drinking and driving ad. However, in the con- dition calling for the presence of an attention-engaging device, it appeared at the beginning of the commercial, whereas it was placed at the end in the alternative condition.
Sample, Procedure and Measures
The experimental sample was comprised of 264 un- dergraduate students, with 134 participants viewing the Mobil 1 commercial and the remaining 130 as- signed to the drinking and driving ad conditions. Treatment conditions were randomly assigned to ses- sions.
Upon the subjects' arrival at the experimental lo- cation, a researcher informed them that the purpose of the study was to assess viewer enjoyment of some recently aired television programs. Subjects then viewed one of the experimental videotapes in a dark- ened room. Two minutes after the end of the test ad, the video monitors were switched off, and the re- searcher distributed the appropriate version (Mobil 1 or drinking and driving) of the measurement instru- ment. After completing the questionnaire, subjects were thanked for their participation and dismissed.
The measurement instrument assessed commercial- relevant cognitive response activity using a form which required subjects to list each thought separately.
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
March 1993 53
Subsequent analysis was thus based upon thought totals provided by the subjects. The instrument also provided checks of manipulations and other experi- mental aspects warranting validation (program in- volvement, recall of and response to attentional de- vices, intrinsic involvement in topic of the test ad and recall), and tested for program mood state and hy- pothesis guessing.
The scale measuring intrinsic involvement in the ad topic employed four seven-point semantic differ- ential items scaled +3 to -3 (reverse coded as needed to attach the positive score to positive anchors) - important/unimportant, irrelevant to me/relevant to me, means a lot to me/means nothing to me, not needed/needed (Cronbach alpha of .87 for Mobil 1 and .85 for drinking and driving). For this and other multiple-item scales, scale means were employed in later analysis. Other manipulation checks included Zaichkowsky's (1985) twenty-item involvement scale used to measure program involvement (alpha = .96), a dichotomous measure of recognition memory for the attentional device (recalled/not recalled), and a seven-point agree/disagree scale assessing whether the device 'took my attention away from the program." Subjects responded to two semantic differential measures of program mood - uplifting/depressing and positive/negative (correlation = .72).
When asked at the end of the experiment to de- scribe what they believed to be the purpose of the study, no one made any connection between program and commercial. Most subjects tended either to be- lieve the cover story or to view the experiment as a conventional advertising test, independent of program context. The latter perception was not viewed as a threat to internal validity, since the impact of such a belief should be the enhanced processing of commer- cials across all conditions, thus minimizing differ- ences between experimental cells and rendering hy- pothesis tests more conservative than in an environ- ment completely uncontaminated by hypothesis guess- ing.
Results
Manipulation Checks The program-involvement scale yielded a signifi-
cantly higher score in the high program involvement
( = .64 on a scale of -3 to +3) than in the low program involvement (x = .38) condition (t = 2.12, p < .05). The significance of this difference, based as it is on a post hoc self-report measure which would not be expected to capture existing variance as effectively as the im-
mediate and more sensitive response time measure employed in the pretest, provides further support for the manipulation of program involvement.
The demonstration of ad involvement differences is important in establishing the domain of the respective ads as vehicles for testing high and low ad-involvement
hypotheses. The topics of the two ads differed in in- volvement, with the four-item intrinsic involvement
scale yielding a higher mean score for the drinking and driving ad (3x = 1.84) than for the Mobil 1 com-
mercial (7x = 1.36; t = -3.37, p < .005). This is not a manipulation check per se, since ad involvement is
not a treatment factor in the experimental design (analyses are executed separately for each ad). The observed difference in self-report ad involvement scores, though significant, is not large. The picture that emerges on the basis of that measure alone would
seem to be that of a slightly involving ad and a mod- erately involving ad, rather than one low and one
high in involvement. Therefore, subsequent references to "high' and "low" ad involvement should be inter- preted in a relative rather an absolute sense. Addi- tionally, since involvement was but one of the differ- ences between the two ads, unmeasured unique mes- sage characteristics may have exerted some influence on responses. Nevertheless, the objective of employ- ing ads that differ significantly in their involvement was obtained. Experiment II, reported hereafter, was designed to isolate ad-involvement effects from those of other message characteristics.
Both attentional mechanisms were recalled by most subjects in the device-present condition (72.3 percent for Mobil 1, 63.2 percent for drinking and driving), and they neither agreed nor disagreed with a state- ment asking whether the device took their attention away from the program (mean of 2.95 on a 1-5 scale for the Mobil 1 beeps, and 2.55 for the drinking and driving skeletal crash scene). This result indicates that neither device was sufficiently strong to gener- ate extreme responses on this scale. On the other hand, subjects may have been limited in their ability to precisely identify the source of attentional shifts
via retrospective self-report measures. As noted ear- lier, both devices yielded evidence of validity in the stimulus development and pretesting stages. The fmal experimental results provide evidence of the manipulation's predictive validity.
Cognitive Responses: Qualitative Overview A qualitative overview of the cognitive responses
to the test ads generated across the sample reveals a broad array of response targets, valence and levels of
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
54 Journal of Advertising
Table 1
Mean Thought Totals - Experiment I
Program Attention Device Commercial Involvement
Level Present Absent
Low
Mobil 1 (Hypothesis 1) 5.08 4.44 (Low Ad Involvement) High*
(Hypothesis 2) 4.49 5.49
Low
Drinking and Driving (Hypothesis 3) 5.77 6.00 (High Ad Involvement) High*
(Hypothesis 4) 6.96 5.97
* significant difference between attention device conditions (p < .05)
elaboration. For each ad, the most consistently recur- ring response was one reflecting a general recogni- tion of and agreement with its central message. Ex- amples are:
Mobil 1: "oil won't freeze" and "works well in cold
weather";
Drinking/driving: "drunk drivers kill many innocent people" and
"if you're really their friend you won't let them drive.'
Descriptions of personal behavior and intentions were common responses to both ads:
Mobil 1: "would consider Mobil af- ter seeing this commer- cial,'
"would use if I had a new car and were subject to
extreme cold," "I don't use Mobil 1 and probably never will";
Drinking/driving: 'I have taken keys away," "I have driven drunk and probably will again if I can handle it,"
"never again!"
Both ads also elicited thoughts about the conse- quences of using the promoted product or performing (or failing to perform) the advocated behavior:
Mobil 1: "makes car run better," 'improves gas mileage," "lasts longer";
Drinking/driving: "may save a life," "hurt feelings would result," "I would never forgive myself if a friend was injured or killed because I didn't stop him."
Some counterargumentation occurred in response to both ads:
Mobil 1: "will it be too thin in the summer?"
"nothing was proven,' "how often does it get to 35 below?";
Drinking/driving: "a few drinks won't really impalr one's driving ability,"
"sensational hype on this topic - a little over done."
Some other categories of responses occurred largely in response to one ad or the other. The Mobil 1 ad elicited:
Brand name "Mobil means quality," associations: "Mobil has sponsored
many car races;"
Attribute evaluations: "may be too expensive," "requires a can opener,'
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
March 1993 55
"color makes me feel as if it is cleaner for my car,"
"able to withstand cold because of synthetic properties;"
Ad evaluations or "some nice effects," elaboration of "pretty - rose petals executional elements: shattering,"
"informative,"
"deep voice added
strength and authority."
Viewers of the drinking and driving ad reported:
Elaborative "think of the family," responses "friend in intensive care personalized to from accident,"
themselves or "Randy got fined $100 for others: pulling off the road when
too drunk to drive,"
"I hope someone would confront me if I was drunk;"
Thoughts centered "frequent bars that I don't on alternatives have to drive to," to the advocated "let's take a walk and behavior: check out the girls,"
"get help or call police."
Though multiple content differences exist to which the different types of responses observed for the two ads may be attributed, the greater prevalence of per- sonalized elaborations in response to the drinking and driving ad and the greater attention to executional elements in the Mobil 1 commercial lend at least anecdotal evidence to the claimed involvement differ- ences (Petty and Cacioppo 1986).
Hypothesis Tests As noted earlier, the Mobil 1 ad was used to pro-
vide a test of hypotheses focusing on low ad involve- ment; the drinking and driving ad was employed as the high ad-involvement stimulus. Mean thought to- tals observed in each cell defined by the treatment factors (program involvement and attentional device) are depicted in Table 1. Significant main effects of the treatment variables were not expected, and none were obtained.
Support for Hypotheses 1 and 2 would require a significant interaction between the program involve- ment and attentional device factors in affecting the quantity of ad-relevant thoughts induced by expo- sure to the low involvement ad (Mobil 1). The interac-
tion was significant (F = 4.009, p < .05) and in the predicted direction - a result which lends credence to the total pattern implicit in the combined hypoth- eses. Full support for the individual hypotheses de- pends on finding significant differences between pro- gram involvement levels at each level of the attentional device factor. Under this criterion, Hy- pothesis 1, which predicted greater ad-relevant thought in the presence than in the absence of the attentional device when both ad and program in- volvement are low, was not statistically supported (t = .95, p> .10), despite directionally consistent means. Hypothesis 2 predicted an opposite pattern of results when ad involvement is low and program involve- ment is high. This hypothesis obtained statistical, as well as directional, support (means of 4.49 and 5.49 for presence and absence of attentional device, re- spectively; t = -2.15, p < .05).
Hypotheses 3 and 4 focus on the implication of pro- gram involvement and attentional devices for ads which are involving to the audience. Here, research- ers predicted that the presence of such a device would interfere with ad-relevant thought when program in- volvement is low (Hypothesis 3) and facilitate it when the viewer is more highly involved in the program (Hypothesis 4). Again, the interaction of program in- volvement and attentional device was significant (F = 3.896, p = .05). Results were directionally supportive but did not attain significance in the low program- involvement condition (t = -.38, p > .10), thus failing to confirm Hypothesis 3. Hypothesis 4 was supported, with mean numbers of thoughts 6.96 and 5.97 for device-present and device-absent conditions, respec- tively, when program involvement was high (t = 2.30, p < .05). Hypothesis 4 thus obtained direct statistical support, whereas Hypothesis 3 fell short of confirma- tion despite directionally consistent results. This significant interaction also demonstrates that the attentional device did not operate in the same man- ner to affect attentional processes when placed at the end of the drinking and driving ad ("absent" condi- tion) as when it appeared at the beginning ("present"). (Any such retroactive effect on attention and elabora- tion should have obscured the observed interaction.)
To test whether observed program context effects were driven by differences in program involvement or by program-induced mood states, the program-mood index described earlier was added to the ANOVA model as a covariate. The perceived program mood tone covariate had no significant effect on the depen- dent variable, nor did its inclusion attenuate any of the significant results. Accordingly, it appears that
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
56 Journal of Advertisisng
Table 2
Mean Thought Totals - Experiment 2
Attention Device Program Involvement
Level Present Absent
Low* (Hypothesis 3) 4.54 6.86 High (Hypothesis 4) 4.98 4.56
^ significant difference between attention device conditions
(p < .05)
program context effects observed in Experiment I are attributable to involvement rather than mood.
The hypotheses which guided the first experiment were partially supported by the data. The strongest results were associated with the high program-in- volvement condition. Given the support for Hypoth- eses 2 and 4, it appears that the presence of an at- tention-engaging mechanism successfully interrupts ongoing program-relevant thought to facilitate ad elaboration for a commercial of high personal rel- evance. Alternatively, when ad involvement is low, the viewer, who lacks intrinsic motivation to elaborate upon the commercial, tends to be further distracted by the insertion of a dramatic device in a highly in- volving program context and, therefore, engages in less message-relevant thought than in the absence of such a device. The data provided an inadequate basis for making inferences about the effect of attentional devices when program involvement is low since the pattern of means across conditions, though directionally consistent with the hypotheses, did not reach conventional levels of significance.
The generalization of the results to other adver- tisements in highly involving program contexts is constrained by the design limitations of the first ex- periment. The use of different commercials and attentional devices, while providing an externally valid manipulation of involvement by ensuring sub- stantial variation in intrinsic sources of personal rel- evance, leaves open the possibility that observed dif- ferences in viewer responses may be attributable to ad or attentional device characteristics other than involvement. For example, the beeps appended to the Mobil 1 commercial may have been innately more distracting than the skeletal hand/screeching brakes image accompanying the drinking and driving ad. The latter device may also have been more likely to enhance processing by virtue of its relevance to the
ensuing message. These considerations motivated Experiment II, which employed a partial replication of the first study in an attempt to ascertain whether inducing high extrinsic involvement in the process- ing of the Mobil 1 ad would yield results comparable to those previously observed for the more highly in- volving drinking and driving commercial.
Experiment II
Objectives and Research Design
As a partial replication of the foregoing study, Ex- periment II was designed to provide a further test of Hypotheses 3 and 4 which address conditions of high ad involvement. It also tested the extent to which differences in subjects' responses to the Mobil 1 and drinking-and-driving ads in the prior study are at- tributable to the related ad involvement levels, as opposed to other unique message characteristics.
To accomplish these objectives, a single commercial - the Mobil 1 ad employed in the first experiment was used. However, in this experiment, high extrinsic involvement in the ad was induced via oral instruc-
tions. Otherwise, the design remained the same as that employed in Experiment I - a two (program involvement) by two (attentional device) factorial - and used the same program positions and attentional stimulus.
Since viewers may be motivated to process an ad in some depth because of an inherent personal interest in its topic or because of some external situational inducement, an inducement of extrinsic involvement may be expected to yield information processing lev- els comparable to those resulting from intrinsic prod- uct involvement. For example, an otherwise uninvolved viewer who has just been asked to accom- pany a friend on a shopping trip to purchase a new stereo system (extrinsic involvement) may be as in-
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
March 1993 57
clined to attend to a commercial announcing a sale at an electronics goods retail outlet as a committed au- diophile (intrinsic involvement). This implies that, if the differences between Mobil 1 and drinking-and- driving results in Experiment I are legitimately a function of ad involvement, inducing a high level of extrinsic involvement in the Mobil 1 commercial should generate cognitive response results consistent with those obtained earlier for the drinking-and- driving ad. In order to induce the desired level of extrinsic ad involvement, subjects were told at the beginning of the sessions that they would be asked for their impressions of both the program and the advertisements upon completion of the program seg- ment. This differed from the first study, in which viewers were led to expect questions only about the program.
Sample, Procedure and Measures
The sample for Experiment II was comprised of 133 undergraduate management students who partici- pated for extra credit. The experimental procedure was identical to that employed in Experiment I, except for the extrinsic-involvement induction described in the prior section. The original instrument developed for Mobil 1 conditions in the first experiment was modified by deleting the intrinsic ad involvement items. The free recall measure was used to assess the effectiveness of the extrinsic-involvement induction. (The intrinsic ad involvement scale was incapable of tapping extrinsic processing motivation since the same commercial was employed for both ad involvement conditions.)
Results
Manipulation Checks Manipulations were again successful. Program in-
volvement varied significantly between the high and low conditions (means of .41 and -.09, respectively, t = 2.87, p < .01). Likewise, those in the high program- involvement condition reported a significantly higher level of ongoing thinking about the program during commercial time than did those in the low condition (3.34 and 2.72, respectively; t = 2.84, p < .01). The audio tones employed as an attention-engaging de- vice were recalled by most subjects in the device- present condition (79.1 percent), and the statement that the device took their attention away from the program elicited, on average, a mild level of agreement ( 3.13 on a five-point scale). Finally, the extrinsic-
involvement induction appears to have been succes- ful as evidenced by a higher mean number of message elements recalled in this sample (10.07) than was obtained for the Mobil 1 ad in the prior study (8.61).
Hypothesis Tests Total thought means are displayed in Table 2. In
view of the objectives of Experiment II, discussion centers on both the implications of the differences between cell means with respect to the hypotheses and on the comparability of the results to those ob- served for the high and low involvement ads in the prior study.
Consistent with the theoretical premise for Hy- potheses 3 and 4, the interaction between the program
involvement and attentional device factors was sig- nificant (F = 8.254, p < .01), with no significant main effects of the treatment variables. This pattern of results corresponds directionally with those of the drinking-and-driving ad in the first study, and is es- sentially opposite those observed earlier for the Mobil 1 commercial in the absence of the extrinsic-involve- ment induction. Hypothesis 3, which failed to obtain support in the first study despite directionally con-
sistent results, was supported in this instance. Those subjects viewing the ad preceded by an attentional
device in the low program-involvement condition generated fewer ad-relevant thoughts than did those seeing the ad without the audio tones (means of 4.54 and 6.86, respectively; t = -2.35, p < .05). Hypothesis 4, involving the high program-involvement condition, failed to obtain support in Experiment II. There was a nonsignificant difference in means in the hypoth- esized direction (t = 1.04, p > .10).
The program mood covariate was again nonsignifi- cant and failed to result in the attenuation of any of the significant results described above. Finally, the hypothesis guessing measure did not reveal any de- mand effects.
Discussion
Inferences
At the most basic level, the results of the two ex- periments demonstrate clearly that program in- volvement, a construct which most prior literature has assumed to operate largely independently of ad or viewer characteristics, must be viewed in the con- text of the ad's inherent attention-engaging capacity and the viewer's intrinsic or extrinsic involvement in the message. It was theorized that viewer processing
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
58-- -----Journal of Advertising
of commercial messages would be an interactive func- tion of program, message and viewer characteristics. That expectation was confirmed.
While further research is needed to clarify some ambiguities, there is sufficient consistency between the two studies to assert that the elaboration of com- mercial message content can benefit from, or be hin- dered by, program involvement. The advantage of placing a commercial in a highly involving program context is that it reaches a viewer whose attentional resources are already activated. In a scenario where motivation to process is low and the ad does not employ a dramatic device to immediately draw attention to itself (Hypothesis 2), program involvement serves to heighten viewer response. Alternatively, if the ad is one in which the viewer is already interested, high program involvement may delay the transfer of attentional focus from the program to the commercial in the absence of an attentional device, thereby re- ducing the amount of message elaboration (Hypoth- eses 3 and 4). The attention-engaging device appears to facilitate ad elaboration in situations in which program-relevant thought is temporarily distracting viewers from focusing their attention on an advertis- ing message of inherent or situational interest (Hy- pothesis 4). Otherwise, it appears to serve only as a further distractor (Hypotheses 2 and 3).
The demonstration in Experiment II of a crossover interaction between the two experimental factors (pro- gram involvement and attentional device), which is directionally consistent with that obtained for the high-involvement ad and opposite that observed for the low-involvement ad in the first study, effectively argues against a dominant role for message differences other than involvement in accounting for these results. Likewise, the consistent failure of the program mood covariate to attain significance or to attenuate pro- gram involvement-attention device interactions eliminates mood as an alternative explanation for program context effects in these studies.
Implications
The advertiser who is hoping to find in these re- sults a single, universally optimal approach to the program involvement issue will be disappointed. Rather, the results of this research suggest that a variety of strategies, depending on the extent to which an ad's effectiveness requires active cognitive pro- cessing, the level of viewer motivation to process the ad, and the nature of the program environment in which the ad is placed, may be necessary in order to
induce a desired level of message rehearsal and elabo- ration.
If ad involvement is low, high program involvement may induce the desired state of cognitive activity. Alternatively, an attention-engaging device may per- form the same function as high program involvement. To the extent possible, however, it would be desirable to avoid the combination of multiple antecedents of cognitive arousal (e.g., high program involvement and an attentional device), since they may interact to distract viewers from commercial-relevant thought.
When ad involvement is high (either intrinsically or because of extrinsic inducement), a low-elaboration program position appears to be desirable. If the pro- gram environment is likely to elicit high levels of viewer elaboration, a device which immediately and dramatically calls attention to the product or issue in which the viewer is already highly involved may counter the program's distraction potential.
The use of a public service announcement (the drinking and driving ad) as one of the stimuli in the first experiment demonstrates the applicability of the results to this domain as well as to paid advertising. While in the present instance a highly involving ad- vocacy message was employed, many public service announcements (PSAs) reach audiences less motivated to attend to and elaborate upon them. Implicit in the often-erratic scheduling of PSAs is a lack of control on the part of the sponsoring organization over both the audience reached and the program context. Tele- vision PSA directors can, however, provide an adver- tiser with a sense of programs or times which their stations tend to favor for the airing of such a message. Any such information on likely program and audience characteristics can reduce the publicist's uncertainty with respect to the advisability of opening a PSA with a dramatic attentional device.
Limitations
While the two experiments provide reasonable evi- dence for the hypothesized relationships between pro- gram and message involvement and the attention- engaging capacity of an ad, they share some common limitations. The difference between the two attentional devices employed in Experiment 1 with respect to their inherent relevance to the message which fol- lowed (beeps irrelevant to Mobil 1, skeletal hand and car crash image relevant to drinking and driving) may partially account for the differences in results between the two experiments, though the experimen- tal design precludes a direct empirical assessment of
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
March 1993 59
the impact of this stimulus characteristic. However, the reversal in direction of the Mobil 1 cross-over
interaction between Experiments I and II, which make it directionally consistent with the drinking and driv- ing ad when response involvement is induced (in spite of the continued use of beeps as the attentional device), argues against a major role of device relevance in explaining the differences observed in the first study. The results thus appear reasonably robust to sub- stantial vanrations in the character of the attentional device. Additionally, the placement of the skeletal hand/car crash device at the end of the drinking and driving ad in the first experiment's "device absent" condition, while ruling out time and content differ- ences as confounding variables, potentially raises a question of post-message attention arousal effects. As noted earlier, however, if any such effects occurred they were not of sufflcient strength to substantially obscure the expected differences between "present" and "absent" conditions.
A reviewer noted that effects attributed to program involvement cannot be unambiguously isolated from the more general issue of commercial position, since the same program positions were employed in both studies, with the low involvement positions preceding the high involvement commercial break. Aside from program involvement, the most obvious position characteristics which would operate in this context would seem to be mood or tedium. As noted earlier, mood was assessed and found not to be a problem. Both pretest response latency indicators and the program involvement manipulation check provide support for the conclusion that the high involvement position is higher in involvement than the low in- volvement position. These and the cognitive response
data argue against the tedium alternative. Thus, it does not seem that position differences can account for the observed results.
The program involvement effects observed in this study do not, at this point, have the advantage of replication across varied program contexts. The use of a single program, while effectively minimizing non- involvement based program differences between con- ditions, does not allow complete confidence in the generalization of results to other programs sharing comparable involvement differences. Likewise, the ad involvement manipulations used in Experiment I pre- clude the separation of effects of advertising involve- ment from other idiosyncratic differences inherent in the two test ads. While the reversal of the crossover interaction in Experiment II effectively argues against a dominant role of such idiosyncratic differences, the
lack of perfect replication suggests that some unex- plained effects remain.
The use of two totally different ads and attentional devices in Experiment I (replication rather than ma- nipulation), and the singular focus on high extrinsic ad involvement in Experiment II, precluded a statis- tical assessment of the impact of intrinsic or extrinsic ad involvement and its interaction with the two ex-
perimental factors. Additionally, the exclusive use of cognitive response measures, while appropriate for the testing of our process-oriented hypotheses, pre- cludes the generalization of our results to such out- come variables as brand awareness, recall, attitude, and intention.
Future Research
A need exists for more extensive work to further elucidate the effects of program involvement on com- mercial processing. The present research could benefit from modified replications, employing stronger ma- nipulations, different program contexts, a broader array of process and outcome dependent variables to more fully tap audience response hierarchies, coun- terbalancing of high and low program involvement positions, and the experimental manipulation of ad involvement. Additional work is needed to clarify the impact of attention-engaging devices. Extensions of this research could focus on investigation of devices differing on such dimensions as their level of attentional stimulation, their relevance to the mes- sage, and their placement within the message. A more comprehensive approach to the investigation of ad involvement results, employing designs which allow that construct to serve as a treatment or blocking factor in order to fully observe its interaction with program and ad characteristics, would be beneficial. To facilitate the effective implementation of these results in the public service arena, research into the scheduling paradigms employed for the airing of PSAs is also needed.
References
Anand, Punam and Brian Sternthal (1990), 'Ease of Message Processing as a Moderator of Repetition Effects in Advertising," Journal of Marketing Research, 27 (August), 345-353.
Batra, Rajeev and Michael L. Ray (1983), "Operationalizing In- volvement as Depth and Quality of Cognitive Response," in
Advances in Consumer Research, Vol. X, Richard Bagozzi and Alice Tybout, eds., Ann Arbor, MI: Association for Consumer Research, 309-313.
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
60 Journal of Adverti'sing
and (1986), 'Affective Responses Medi- ating Acceptance of Advertising,' Journal of Consumer Re- search, 13 (September), 234-249.
Britton, Bruce K., Shawn M. Glynn, Bonnie J. Meyer, and M.J. Penland (1982),'Effects of Text Structure on Use of Cognitive Capacity During Reading,' Journal of Educational Psychol- ogy, 74 (February), 51-61.
and Alan G. Sawyer (1983), 'Effects of Involvement and Message Content on Information-Processing Intensity,' in Information ProcessingResearch in Advertising, Richard J. Harris, ed., Hillsdale, NJ: Laurence Erlbaum Associates, 43- 64.
Burkrant, Robert E.(1960), Arousal and Curiosity, New York: McGraw-Hill.
, H. Rao Unnava, and Kenneth R. Lord (1992), "The Effects of Programming Induced Mood States on Memory for Commercial Information,' working paper, The Ohio State University College of Business.
Bryant, Jennings and Paul W. Comisky (1978), 'Te Effect of Positioning a Message within Differentially Cognitively In- volving Portions of a Television Segment on Recall of the Message,' Human Communication Research, 5 (Fall), 63-75.
Celsi, Richard L. and Jerry C. Olson (1988), 'The Role of Involve- ment in Attention and Comprehension Processes,' Journal of Consumer Research, 15 (September), 210-224.
Chaiken, Shelly, Akiva Liberman and Alice H. Eagly (1989), "Heuristic and Systematic Information Processing Within and Beyond the Persuasion Context,' in Unintended Thought: Limits of Awareness, Intention and Control, James S. Uleman and John A. Borgh, eds., New York: Guilford, 212-252.
Fiske, Susan T. and Mark A. Pavelchak (1986), "Category-Based Versus Piecemeal-Based Affective Responses: Developments in Schema-Triggered Affect,' in The Handbooks of Motivation and Cognition: Foundations of Social Behavior, Richard M. Sorrentino and E. Tory Higgins, eds., New York: Guilford, 167-203.
Gardner, Meryl Paula (1985), "Mood States and Consumer Behavior: A Critical Review,' Journal of Consumer Research, 12 (De- cember), 281-300.
Goldberg, Marvin E. and Gerald J. Gomn (1987), "Happy and Sad TV Programs: How They Affect Reactions to Commercials,' Journal of Consumer Research, 14 (December), 387-403.
Greenwald, Anthony G. (1968), "Cognitive Learning, Cognitive Response to Persuasion, and Attitude Change,' in Psycho- koical Foundations of Attitudes, Anthony G. Greenwald, Timothy C. Brock, Thomas M. Ostrom, eds., New York: Aca- demic Press, 147-170.
Kahneman, Daniel (1973), Attention and Effort, Englewood Cliffs, NJ: Prentice Hall.
Kennedy, John R. (1971), "How Program Environment Affects TV Commercials,' Journal of Advertising Research, 11 (Febru- ary), 33-38.
Krugman, Herbert E. (1983), "Television Program Interest and Commercial Interruption: Are Commercials on Interesting Programs Less Effective?" Journal ofAdvertising Research, 23 (February/March), 21-23.
Lord, Kenneth R. and Robert E. Burnkrant (1988a), 'Television Program Elaboration Effects on Commercial Processing,' in Advances in Consumer Research, Vol. XV, Michael J. Houston, ed., Provo, UT: Association for Consumer Research, 213-218.
and - (1988b), "Commercial Processing Ef- ficiency: A Program Elaboration Model," in 1988 AMA Educa- tors' Proceedings: Efficiency and Effectiveness in Marketing, Gary Frazier, Charles Ingene, David Aaker, Avijit Ghosh, Tom Kinnear, Sidney Levy, Richard Staelin, John Summers, eds., Chicago: American Marketing Association, 112-116.
and Robert S. Owen (1989), 'An Ex- perimental Comparison of Self-Report and Response Time Measures of Information Processing," in AMA Educators'Pro- ceedirgs: Enhancing Knowledge Development in Marketing, Paul Bloom, Russ Winer, Harold Kassarjian, Debra Scammon, Bart Weitz, Robert Spekman, Vijay Mahajan, Michael Levy, eds., Chicago: American Marketing Association, 196-200.
Maclnnis, Deborah T., Christine Moorman, and Bernard J. Jaworski (1991), 'Enhancing and Measuring Consumers' Motivation, Opportunity, and Ability to Process Brand Information from Ads,> Journal of Marketing, 55 (October), 32-53.
Miller, George A. (1956), 'The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing In- formation," Psychological Review, 62 (March), 81-97.
Moore, Danny L., Douglas Hausknecht, and Kanchana Thamodaran (1986), 'Time Compression, Response Opportunity, and Per- suasion," Journal of Consumer Research, 13 (June), 85-99.
Murry, John P. Jr., John L. Lastovicka, and Surendra N. Singh (1992), 'Feeling and Liking Responses to Television Programs: An Examination of Two Explanations for Media-Context Ef- fects," Journal of Consumer Research, 18 (March), 441-451.
Nord, Walter R. and J. Paul Peter (1980), "A Behavior Modification Perspective on Marketing,"Journal of Marketing, 44 (Spring), 36-47.
Pavelchak, Mark A., John H. Antil, and James M. Munch (1988), "The Super Bowl: An Investigation into The Relationship among Program Context, Emotional Experience, and Ad Re- call," Journal of Consumer Research, 15 (December), 360-367.
Peter, J. Paul and Jerry C. Olson (1990), Consumer Behavior and Marketing Strategy, Homewood, IL: Richard D. Irwin.
Petty, Richard and John Cacioppo (1986), 'The Elaboration Like- lihood Model of Persuasion," in Advances in Experimental Social Psychology, Vol. XIX, Leonard Berkowitz, ed., New York: Academic Press, 123-205.
Schumann, David W. (1986), "Program Impact on Attitude Toward TV Commercials," in Proceedings of the Division of Consumer Psychology, Joel G. Saegert, ed., Washington, D.C.: American Psychological Association, 67-73.
Simon, Herbert A. (1974), "How Big is a Chunk?" Science, 183 (February), 482-488.
Soldow, Gary and Victor Principe (1981), 'Response to Commercials as a Function of Program Context," Journal of Advertising Research, 21 (April), 59-65.
Thorson, Esther, Byron Reeves, Joan Schleuder, Annie Lang, and Michael L. Rothschild (1985), "Effects of Program Context on the Processing of Television Commercials,"in Proceedings of the American Academy of Advertising, Nancy Stephens, ed., Tempe, AR: American Academy of Advertising.
Wright, Peter L. (1973), "The Cognitive Processes Mediating the Acceptance of Advertising,"Journal of Marketing Research, 10 (February), 53-62.
This content downloaded from 131.247.112.3 on Sun, 25 Mar 2018 15:15:45 UTC All use subject to http://about.jstor.org/terms
- Chap Review
- Jamieson - Good Intuition_or_Fear
- Driver - Integrative Complexity
- Contents
- image 1
- image 2
- image 3
- image 4
- image 5
- image 6
- image 7
- image 8
- image 9
- image 10
- image 11
- image 12
- image 13
- image 14
- Lowenstein - Out of Control
- Salvia - Attractiveness
- Ambady - Half a Minute
- Lord - Advertising
- Contents
- [47]
- 48
- 49
- 50
- 51
- 52
- 53
- 54
- 55
- 56
- 57
- 58
- 59
- 60