You are on page 1of 8

This article was downloaded by: [Umea University Library]

On: 31 July 2010


Access details: Access Details: [subscription number 781079107]
Publisher Psychology Press
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-
41 Mortimer Street, London W1T 3JH, UK

Psychological Inquiry
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t775648164

Chaos: On Making a Convincing Case for Social Psychology


Abraham Tesser; Robert McMillen; Jon Collins

Online publication date: 16 December 2009

To cite this Article Tesser, Abraham , McMillen, Robert and Collins, Jon(1997) 'Chaos: On Making a Convincing Case for
Social Psychology', Psychological Inquiry, 8: 2, 137 — 143
To link to this Article: DOI: 10.1207/s15327965pli0802_10
URL: http://dx.doi.org/10.1207/s15327965pli0802_10

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
COMMENTARIES

comes would be coded as losses than would be the case Note


for lower values of this payoff. What would be the net
effect of increasing this parameter? The answer to this David M. Messick, Kellogg Graduate School of
question was unclear until we developed a model that Management, 356 Leverone Hall, Northwestern Uni-
allowed us to make quantitative predictions. Our intui- versity, 2001 Sheridan Road, Evanston, IL
tions were no longer reliable. In this case, as the simu- 60208-20 11.
lation data showed, cooperation decreased modestly as
the temptation increased, but the process by which this
occurred was completely different from that presumed References
in the individualisticmodel. Understanding how coop-
eration levels change in this simulated environment is Allison, G. T. (1971). Essence of decision. Boston: Little, Brown.
a far more coinplex affair than simply understanding Allison, S. T., & Messick, D. M. (1985). The group attribution er-
ror. Journal of Experimental Social Psychology, 21, 563-579.
the choice mechanisms of a single subject. Cartwright, D., & Zander, A. (Eds.). (1953). Group dynamics.
The emergence (or reemergence) of a dynamical New York: Harper & Row.
social psychology, as outlined by Vallacher and Krech, D., & Crutchfield, R. S. (1948). Theory andproblems of so-
Nowak, challenges social psychologists to revisit ques- cial psychology. New York: McGraw-Hill.
tions about the nature of social phenomena and the Lewin, K. (1935). A dynamic theory ofpersonality. New York:
McGraw-Hill.
relation between explanations in terms of groups and in Messick, D. M., & Liebrand, W. B. G. (1995). Individual heuris-
terms of individuals. Discovering the connections be- tics and the dynamics of cooperation in large groups. Psycho-
Downloaded By: [Umea University Library] At: 08:42 31 July 2010

tween these levels of analysis may be one of the most logical Review, 102, 131-145.
exciting and elusive problems to be addressed in the Miller, N. E. (1944). Experimental studies of conflict. In J. McV.
future. Although there are new mathematical theories Hunt (Ed.), Personality and the behavior disorders @p.
431-465). New York: Ronald Press.
that will provide useful tools in this pursuit, it would be Nowak, A., Lewenstein, M., & Szamrej, J. (1993, September). So-
a pity to ignor~ethe wisdom of our predecessors of 50 cial transitions occur through bubbles. Scientzjtc American
years ago when they developed the original dynamical [Polish version], 12, 16-25.
social psychology.

Chaos: On Making a Convincing Case for Social Psychology


Abraham Tesser, Robert McMillen, and Jon Collins
Department of Psychology
University of Georgia

Vallacher and Nowak have done psychology in gen- understanding of social psychological phenomena. But
eral and social psychology in particular a great service they do so only to the extent that they have precise
in writing the target article for this issue. Since the meaning within some particular, developed formula-
1980s,the physical sciences, such as physics, chemistry tion. Vallacher and Nowak, in their book (Vallacher &
and meteorology, have benefited by analyzing in terms Nowak, 1994) and in the target article, have taken us a
of nonlinear dynamical systems phenomena that ap- long way toward a better understanding of these con-
peared too complex for prediction. The understanding cepts and toward the application of these concepts to
of biological systems, particularly ecological systems, social psychological phenomena.
has also benefited substantiallyfrom this kind of analy- We too are taken with the potential of this approach.
sis. Moreover, with the publication of Gleick's (1987) However, we believe that its adoption has some pro-
Chaos, the general public became aware of this new found implications for the paradigmatic epistemology
"science," and there has been speculation about its of social psychology. Although some of the implica-
application in almost every field. Much of that specu- tions of nonlinear dynamics can be dealt with by patch-
lation is based on only a vague understanding of the ing, changing the emphases, and increasing the sophis-
technical work. There is a widespread intuitive under- tication of how we currently collect data, other
standing and attraction to concepts such as chaos, com- implications require that we abandon some bedrock
plexity, dynamics, and emergent order. These are sexy assumptions and embrace new ways of testing the
words, and they hold promise for an increment in our "truth" of our theories. In this response we review
COMMENTARIES

several implications of the nonlinear dynamics ap- Hysteresis


proach and describe what we believe are the necessary
methodological and epistemological accommodations Hysteresis is a "signature phenomenon of nonlinear
necessary to pursue those implications. relations." Hysteresis is present if for some values of X
(the cause), Y (the effect) can take on two values. Which
of these two values is predicted depends on the history
Paradigmatic Social Psychological of the variables. Evaluating a behavioral system for the
Bedrock presence of hysteresis presents no special challenge to
experimental social psychologists. Indeed, we have
Contemporary social psychology is very much a studied such systems for years. The primacy effect in
prototypical science. It develops theories about causes, impression formation is an obvious example. The im-
mediating processes, and effects. Depending on the pression of another will be more positive if the very
investigators' preferences, the theoretical causes and same descriptors on which it is based are presented in
effects are more or less abstract, but testing the theory a positive-negative order than if they are presented in
always involves translating the variables into opera- a negative-positive order (Asch, 1946). The hysteresis
tional definitions that involve the manipulation or hypothesis invites us to examine more levels of the
measurement of a real system, that is, person(s). The independent variable than we typically do (e.g., Tesser
theory is strengthened if the operational variables relate & Achee, 1994), but very simple designs that take
to one another as predicted, it is weakened if they do account of the history of the system will suffice.
Downloaded By: [Umea University Library] At: 08:42 31 July 2010

not.
Usually the effect or dependent variable is measured
at one point in time. (This is particularly the case where Discontinuous Change
the random assignment of subjects to conditions is
possible.) For any level or any single combination of One of the simplest forms of nonlinearity is discon-
levels of an independent variable(s) there is usually a tinuous changes. Large changes in X across most of its
single predicted value of the dependent variable. The range results in a small or no change in Y. There is,
latter two points describe normative practice, not bed- however, a small region in X to which Y is sensitive,
rock assumptions. What is bedrock is that real-world and a relatively small change in X in this region results
observations must take place in testing a theory, logical in a very large change in Y. This kind of phenomenon
(theoretical) coherence does not substitute for an em- also has a history of study in psychology under the name
pirical test. Finally, somewhere between normative of threshold phenomena, step functions, and binary
practice and bedrock assumption is the generally un- responses. For example, nerve cell firing is a threshold
mentioned assumption that the cause and the effect are phenomenon, item response models in test theory are
not the same variable. Research with nonlinear dynam- sometimes based on step functions, and the personal
ics puts stress on all these assumptions: the normative, constructs by which we negotiate the interpersonal
the unmentioned, and the bedrock. world are binary, according to Kelly (1963).
In spite of this history, most empirical work in social
psychology appears insensitive to the possibility of
On Evaluating Some Nonlinear discontinuity. Although there has been a recent upturn
Implications in the use of diary studies with multiple measures over
time, most of our studies involve few levels of the
Vallacher and Nowak present a number of implica- independent variable, between-subjects designs, and
tions of nonlinear systems for social psychology. We aggregation of data for evaluation. If we are to show
review some of these implications and focus on what it sensitivity to the possibility of discontinuous functions,
would take to evaluate their scientific utility. For exam- each of these practices will have to be reevaluated.
ple, what kind of "tests" are necessary for us to increase Obviously, the greater the number of levels of the
our confidence that some phenomena show extreme independent variable, the more informative the research
sensitivity to initial conditions, that a change process is can be with regard to shape. There is, however, a more
discontinuous, and so forth. We ordered the implica- subtle point to be made that has to do with the range of
tions on which we focus. We start with implications that X and just what part of the range is articulated. Suppose
can be tested via conventional means, move on to we measure a child's reading ability every week during
implications that push the envelope on our normative his or her 7th year. We might observe a curve showing
methods, and conclude with implications that appear to smooth, continuous improvements over the year. On the
require basic epistemological revisions. other hand, suppose we measure reading ability once a
COMMENTARIES

year from birth until age 50. We might find that the demands nothing radical. It requires that we pay closer
function is discontinuous:little happening over the first attention to scales of measurement, their range, and
6 or 7 points, big increase over the points from 7 to 11, their articulation; it pushes us to take many measures
and a leveling off after that. Rather than a smooth over time or any other variable of interest, and it alerts
function, the second study would suggest that reading us to the differential strategic consequences of "true"
ability is closer to a discontinuousfunction. This exam- individual differences and errors of measurement.
ple also makes clear that the place in the range where
the scale is articulated will also determine the outcome.
Weekly measurement in Year 46 will show little rela- Fractal Dimensionality
tion between time and reading ability. Finally, these
examples also make clear that whether we regard the One of the aspects of nonlinear dynamics that has
relation to be smooth, discontinuous, or null depends done as much as any other to fire the imagination is
crucially on the scale of measurement, that is, its range fractal dimensionality, introduced by Mandelbrot
and articulation (van Geert, 1991). (1982). By now, all of us have seen beautiful pictures of
If the true relation between X and Y is discontinuous the Mandelbrot set. This is the lacy seashelllparamecium
but there are individual differences in the threshold for figure whose detail can be blown up to reveal more lacy
the "jump," aggregation will obscure a discontinuity seashells, whose details can be blown up to reveal even
even if the scale of X is appropriate. Suppose we more lacy seashells, and so on. No matter how deeply
observe Y at each of multiple levels of X. If the thresh- we probe the detail of the set, it reveals self-similarity
Downloaded By: [Umea University Library] At: 08:42 31 July 2010

old is different for different subjects, each succeeding on smaller and smaller scales of measurement. Nature is
level of X will have a larger proportion of subjects for full of fractal entities that reveal similar morphology
whom the threshold has been exceeded, and the means over several different scales of measurement. Any part
(aggregate statistic) will appear to show a smooth in- of a cloud formation reveals a formation similar to the
crease. This will occur regardless of whether the meas- whole. The branch of a tree reveals a structure similar to
urement across X is made between subjects or within the tree itself, and so forth. A map of a coastline looks
subjects. However, if the measurement is made within roughly similar whether the scale is 100miles tothe inch,
subjects,it is possible to look at the shape of the function 50 miles to the inch, or 1 mile to the inch. Indeed,
for each individual subject (Sidman, 1960) to decide picturing the map of a coastline is a good way to under-
whether the function is continuous or discrete. Thus, if stand the concept of fractional dimensionality and the
there are individual differences in the impact or percep- test for it. As pointed out by Vallacher and Nowak, the
tion of X or in the location of the threshold, one must curve depicting a coastline often has properties of both
take repeated measures within individuals in the appro- a line (one dimension) and a surface (two dimensions).
priate range of the scale to determine if the function is A line has finite length, but the length of our coastline
discontinuous. increases as our scale of measurement decreases, and it
However, th~ereis another side to this analytical coin. reveals more structure, such as coves, bays, and inlets.
Individual differences are not the only problem. If X is Its dimensionality is fractional, that is, more than one
only one of many uncorrelated contributors to Y, or if dimension, but less than two. The discovery that order
Y is measured with substantial error, the discontinuity variables (or dependent variables) have fractal dimen-
will again be obscured. In this circumstance, however, sionality is important because it indicates that the vari-
rather than disaggregation, one of the royal roads to able is part of a chaotic system.
clearer understanding is aggregation. Aggregation over There are ways of determining the dimensionality of
subjects and conceptually equivalent occasions tends to an order variable. Vallacher and Nowak (Vallacher,
cancel out unc~orrelatedcauses and errors of measure- Nowak, & Kaufman, 1994) successfully used an algo-
ment, and what emerges is a much clearer picture of the rithm developed by Grassberger and Procaccia (1983)
true relation (Tesser & Beach, 1995). A good example to estimate the dimensionality in social judgments over
of this strategy is the research on the so-called P300 time. This method estimatesdimensionality by compar-
response in psychophysiology (e.g., Donchin, Karis, ing changes in information (Attneave, 1959) as one
Bashore, Coles, & Gratton, 1986). Only the aggregated systematically controls the previous trial, the previous
data clearly reveal the shape of the curve relating brain two trials, and so forth. In order to estimate dimension-
response over time to the input of "oddball" stimuli. ality, however, one needs many datapoints. Not on the
The nonlinear dynamics approach alerts us to the order of 5 or 6 per subject, but more on the order of 500
very real possibility that some responses may be dis- or 600 per subject. (In order to save space, we follow
continuous. Testing for this possibility pushes the Vallacher et al., 1994, and assume that the subject is the
norms of social psychological research methods but unit of choice.)
COMMENTARIES

Getting this number of datapoints has a number of (e.g., 1988) work on self-evaluation maintenance is
nontrivial implications for how social psychologists do clearly a systems approach: There are three important
their science. First, one must examine a system whose variables, self-relevance,performance relative to other,
temporal dynamics are revealed in a reasonable amount and closeness of other. Each variable is both an inter-
of time. Vallacher et al. (1994) showed that some social active cause and an effect of the other two variables.
judgments can be usefully studied in a single laboratory Self-regulation theories (e.g., Carver & Scheier, 1981)
session. On the other hand, the full richness of close also clearly are systems formulations. And, as Val-
relation dynamics may take years to play themselves lacher and Nowak point out, one of us (e.g., Tesser,
out. Again, scale of measurement is crucial. If we are 1978) has even conducted a research program on the
to stay in the lab and work with relatively fast phenom- internal dynamics of evaluation simply as a result of
ena, we will have to develop methods for capturing lots time and thought within the context of the "usual"
of observations in a short time. The mouse paradigm paradigm. Nevertheless, we argue that a real difference
developed by Vallacher and Nowak seems to be quite between sympathy and practice exists.
promising for studying conscious shifts in judgments Why the gap between sympathy and practice? Psy-
and preferences. Many physiological measures yield chologists are sympathetic to systems-like notions of
continuous data and are related to positive affect, nega- behavior because such notions are intuitively appealing
tive affect, arousal, inhibition, internal versus external in an informal way: Everything is related to everything
information processing, and motivational states. Social else, changes can have unforeseen consequences, small
psychologists interested in temporal dynamics may do causes can result in large changes, and so forth. Al-
Downloaded By: [Umea University Library] At: 08:42 31 July 2010

well to put more effort into learning these systems. though they are intuitively attractive, however, sys-
Finally, diary studies and beeper studies may be de- tems-like models are difficult to deal with formally. A
signed to sample more frequently, to continue sampling variable is a cause and, at the same time, an effect. To
for longer periods of time, or both. One might even understand the system, one has to keep in mind not only
envision subjects equipped with unobtrusive devices bidirectional relations that might go in different direc-
that frequently sample gross motor activity, heart rate, tions, but combinations of relations, and so forth. But
or blood/saliva chemistry during subjects' normal day- we believe the most important reason for systems-like
to-day activities. All of this, however, is a lot of work. research being underrepresented in our journals is that
Therefore, those who are optimistic about the approach we simply do not yet have the methods to adequately
will have to provide particularly compelling theoretical analyze the data from a system. It is true that we have
arguments or empirical examples of the utility of the become more sophisticated in analyzing data from time
approach in order to enlist others. series (e.g., Dorfman & Cacioppo, 1990). It is also true
that structural equation approaches (e.g., Joreskog &
Sorbom, 1985) can, given certain assumptions and un-
Systems Thinking and Internal der certain conditions, be made to yield bidirectional
Dynamics causal coefficients. It is also true that these approaches
encourage systems thinking. But, for the most part,
The nonlinear dynamics approach puts a strong em- these systems tend to be used for additive combinations
phasis on systems thinking. One might characterize the of effects and linear relations among variables.
traditional approach in social psychology to be sympa- In sum, traditional social psychology is open to the
thetic to this point of view but to have been relatively emphasis that Vallacher and Nowak put on systems-like
indifferent to it in practice. The usual practice has been thinking. Indeed, their target article provides an impor-
to think of phenomena in unidirectional terms, that is, tant step toward making systems-like thinking accessi-
X causes Y, and to pay relatively little attention to the ble. However, methods for evaluating empirical data in
notion that Y may also be a cause of X or that Y may comprehensive systems-like terms have not yet been
have some internal dynamics of its own. sufficiently developed.
It is worth noting that the difference may not be as
great as it first appears. Bidirectional causality has
occupied a central role in the development of the disci- Chaos and Sensitive Dependence on
pline. Many social psychologists study the impact of Initial Conditions
attitudes on behavior (e.g., Fazio, 1990). Cognitive
dissonance theory, one of the most influential theories We confess that our approach to the characteristics
in the history of the discipline, made us aware of the of nonlinear dynamical systems is slightly disingenu-
reverse causal arrow, namely, that behavior can impact ous. We have discussed each characteristic on its own
attitudes (e.g., Festinger & Carlsmith, 1959). Tesser's terms as if the finding of hysteresis or discontinuities,
COMMENTARIES

for example, were an end in itself. They could very well the nonlinear dynamics approach is a simulation re-
be ends in themselves. Indeed, phenomena with these ported by Nowak, Szamrej, and Latant (1990) that
attributes have been productively studied in social psy- showed that with nonlinear attitude change rules and
chology. But this is not why Vallacher and Nowak aspects of Latant's earlier model of persuasion (e.g.,
introduce them. Vallacher and Nowak introduce each Latant, 1981), random distributions of opinions in a
of the features we talked about because they are symp- community become well organized.
tomatic of systems that have properties that are down- This simulation technique can be made more rigor-
right exotic when it comes to normal social psychology. ous by formally subjecting it to a kind of Turing test
The order variable of achaotic system as it unfolds over (Turing, 1950). To perform the test, one records the
time is simply not predictable. Although there may be computer output and behavior from the human system.
partial predictability, that is, ranges of values that the If a judge cannot tell the computer output from the
system does not take on, it is not attracted to a single point human output, the simulation is an adequate one. The
nor to regular periodic change in points. Indeed, its precise systematic comparison with actual behavior makes this
location at any particular point in time is sensitively test more rigorous than the usual simulation. Note that
dependent on initial conditions. This means that very, very regardless of how similar the simulation is to either our
small differences, differences that are smaller than the intuition or to the actual behavior, the result can be no
error that even precise empirical measures will yield, more than a plausibility test (Bem 1972). It simply tells
cascade over time into very large differences. us that the mechanism underlying the simulation could
Downloaded By: [Umea University Library] At: 08:42 31 July 2010

So, how does one put such a model, a model for produce the behavior. It does not tell us that it is the
which specific predictions even in theory are not possi- mechanism that produced the behavior.
ble, to an emlpirical test? It is here that some of the A third possibility is what we call behavioral boot-
bedrock assumptions of normal science and the non- strapping. This method is well illustrated by the work
linear dynamical approach appear to be at loggerheads. of Vallacher et al. (1994; see also our earlier description
Let us start at the descriptive level. It is certainly of of testing for fractal dimensionality).A behavioral sys-
interest to know that there is variable, aperiodic behav- tem is observed with many measures over time under
ior associated with some system. But one of the corner- one or more starting conditions. Then each individual
stones of empirical science, even at the descriptive protocol is described. It can be described in terms of
level, is reliability, or replication.By definition,chaotic periodicity, mean level, variance, or even dimensional-
systems yield different observations whenever they are ity. The observation of fractional dimensionality, or
observed. Moreover, scientists want to move toward quasiperiodicity, and so forth suggests a dynamical
prediction, control, and, ultimately, understanding. system at work. But it tells us precious little about the
Tests of understanding come from theories that tell the system. Although we may learn the dimensionality of
scientist what to tweak and what the result of tweaking the system, we learn nothing about what those dimen-
the system will be. Again, even in principle, such tests sions may be or how they interact with one another. It
about specific values are not possible. is a little like having only the eigenvalues from a factor
analysis. However, in the factor analysis case, at least
one knows that the factors combine additively. The
Is the Situation Hopeless? most realistic expectation from behavioral bootstrap-
ping is the vague demonstration that nonlinear dynam-
The situation is not hopeless. Nonlinear dynamical ics of some sort may be at work. Although this is an
systems have advanced the physical and biological acceptable starting point for empirical work, one hopes
sciences, and those disciplines are certainly no less that better specified models will follow.
"scientific" than is social psychology. There are more From our perspective, some of the applications of
or less rigorous ways in which these models can be nonlinear dynamics to meteorology and biology repre-
scientifically productive. The first is not data driven at sent the most rigorous approach possible. We have yet
all. Computer simulations (or simple demonstrations of to see such an application in social psychology. For
the complexity simple equations imply) can be quite example, the logistic equation, described by Vallacher
persuasive. In the usual case, the result of the simulation and Nowak, had a history of empirical validation as a
is compared to our intuitive understanding of the phe- model of population growth since before the discovery
nomenon. If it resonates with our intuition, the model by May (1973) of its more exotic chaotic properties.
is given some credence. Although this kind of approach Ecologists successfully used the model to make predic-
is not used very often in mainstream social psychology, tions about equilibrium population size on the basis of
it is a mainstay in the field of artificial intelligence. One initial population size and the carrying capacity of the
of the most persuasive demonstrations of the utility of environment. There were, of course, anomalous cases
COMMENTARIES

in which populations, instead of settling down to the ence Foundation Grant SBR 9121276, and National
predicted level, oscillated between two values or even Science Foundation Grant SBR-95 11385. We also are
fluctuated wildly. The realization that the equation grateful to John Achee, Jay Allen, Steven R. W. Beach,
could also account for this kind of behavior simply and Dan Whitaker for their comments on a preliminary
increased the field's confidence in the formulation draft.
Lorenz (1993) had a working simulation of the Abraham Tesser, Department of Psychology, Uni-
weather before discovering the model's sensitive de- versity of Georgia, Athens, GA 30602.
pendence on initial conditions. The model was, of
course, better at predicting the actual weather in the References
short term than in the long term. The model's success
at predicting specific empirical outcomes provides Asch, S. E. (1946). Forming impressions of personality. Journal of
some validity. The fact that it is also sensitively depend- Abnormal and Social Psychology, 41,258-290.
Attneave, F. (1959). Applications of information theory to psychol-
ent on initial conditions must increase our confidence ogy: A summary of basic concepts, methods, and results. New
in the model because this sensitivity implies that the York: Holt.
model will do less well at predicting temporally re- Bem, D. J. (1972). Self-perceptiontheory. In L. Berkowitz (Ed.),
moved weather events. Advances in experimental social psychology (Vol. 6, pp.
Note that the last two examples includeelementsthat 2-62). New York: Academic.
Carver, C. S., & Scheier, M. F. (1981). Attention and self-regula-
map nicely onto the standard paradigm. First, the equa- tion: A control-theory approach to human behavior. New
tions themselves come from a thorough understanding York: Springer-Verlag.
Downloaded By: [Umea University Library] At: 08:42 31 July 2010

of the phenomena in question. There is a one-to-one, Donchin, E., Karis, D., Bashore, T. R., Coles, M. G. H., & Gratton,
theoretical correspondencebetween the variables in the G. (1986). Cognitive psychophysiology and human informa-
equation and the elements of the systems they are tion processing. In M. G. H. Coles, E. Donchin, & S. W. Por-
ges (Eds.), Psychophysiology: System, processes, and applr-
intended to model. The investigators did not start with cations (pp. 244-267). New York: Guilford.
an equation with exotic properties and look for a system Dorfman, D. D., & Cacioppo, J. T. (1990). Waveform moment
into which it could be made to fit. Rather, the equations analysis: Topographical analysis of nonrhythrnic wave forms.
grew out of their knowledge of the system and happened In J. T. Cacioppo & L. Tassinary (Eds.), Principles ofpsycho-
also to have exotic properties. From an epistemological physiology (pp. 661-707). Cambridge, England: Cambridge
University Press.
point of view, the sequence is irrelevant,but we suspect Fazio, R. H. (1990). Multiple processes by which attitudes guide
that the sequence is crucial to the psychology of devel- behavior: The MODE model as an integrative frame work. In
oping a successful, rigorously testable model. In both M. P. Zanna (Ed.), Advances in experimental socialpsychol-
cases, an interaction between theory and data are in- ogy. (Vol. 23, pp. 75-109). San Diego, CA: Academic.
volved. Although there are aspects of the logistic equa- Festinger, L., & Carlsmith, J. M. (1959). Cognitive consequences
of forced compliance.Journal of Abnormal and Social Psy-
tion and the Lorenz equations that produce nonrepro- chology, 58,203-21 1 .
ducible results, there are also aspects that make specific Gleick, J. (1987). Chaos: The d i n g of a new science. New York:
predictions. The fact that the specific predictions are Viking-Penguin.
successful lends greater credence to the aspects of pre- Grassberger, P., & Procaccia, I. (1983). On the characterization of
strange attractors. Physical Revrew Letters, 50, 346-350.
dictions that cannot be made in specific terms. So,
Joreskog, K. G., & Sorbom, D. (1985). LISREL VI user's gurde.
nonlinear dynamical models can be tested rigorously. Mooresville, IN: Scientific Software.
Nonlinear dynamical models hold great intellectual Kelly, G. (1963). A theory ofpersonality. New York: Norton.
interest. They stretch our intuition concerning natural LatanC, B. (1981). The psychology of social impact. American Psy-
phenomena. A plausible case can be made for their chologist, 36, 343-356.
Lorenz, E. N. (1993). The essence of chaos. Seattle: University of
applicability to social psychology. Thanks to Vallacher
Washington Press.
and Nowak, the insights, methods, and procedures Mandelbrot, B. B. (1982). Thefractal geometry of nature. New
needed to carry out this work are more accessible to the York: Freeman.
field. Although changes in epistemology and methods May, R. M. (1973). Stability and complexity in model ecosystem.
may be needed, we along with many of our colleagues Princeton, NJ: Princeton University Press.
Nowak, A., Szamrej, J., & Latan6, B. (1990). From private attitude
hope to see the discipline enriched by the social psy-
to public opinion: A dynamic theory of social impact. Psycho-
chological insights such models may provide. logical Review, 97, 362-376.
Sidman, M. (1960). Tactics of scientiJic research: Evaluating ex-
perimental data inpsychology. New York: Basic Books.
Notes Tesser, A. (1978). Self-generatedattitude change. In L. Berkowitz
(Ed.), Advances in experimental social psychology (Vol. 1 1 ,
pp. 290-338). New York: Academic.
We acknowledge the support of National Institutes Tesser, A. (1988). Toward a self-evaluationmaintenance model of
of Mental Health Grant KO5 MH01233, National Sci- social behavior In L. Berkowitz (Ed.), Advances in expert-
COMMENTARIES

mental socialpsychology (Vol. 21, pp. 181-227). New York: Turing, A. M. (1950). Computing machinery and intelligence.
Academic. Mind, 59, 433-460.
Tesser, A., & Achee, J. (1994). Aggression, love, conformity, and Vallacher, R. R., & Nowak, A. Eds.). (1994). Dynamical system
other social psychological catastrophes. In R. R. Vallacher & in social psychology. San Diego, CA: Academic.
A. Nowak (Eds.), Dynamical sysrem in social psychology Vallacher, R., Nowak, A,, & Kaufman, J. (1994). Intrinsic dynarn-
(pp. 96-109). San Diego, CA: Academic. ics of social judgement. Journal of Personaliiy and Social
Tesser, A., & Beach, S. R. H. (1995). Stressful life events, relation- Psychology, 67.20-34.
ship satisfaction, and negative affect:Some non-linear dynam- van Geert, P. (1991). A dynamic systems model of cognitive and
ics. Unpublished manuscript. University of Georgia, Athens. language growth. Psychological Review, 98, 3-53.

Time and Theory in Social Psychology


Paul van Geert
Department of Psychology
University of Groningen
Downloaded By: [Umea University Library] At: 08:42 31 July 2010

It is hard to deny that Vallacher and Nowak are social psychology that does not meet the criterion
talented architects. Their dynamic-systems-for-social- dynamic. Clearly, this cannot be the meaning of dy-
psychology building is a well-designed and complete namics to which Vallacher and Nowak adhere. Let me
structure. Their article provides an excellent and in- try to make this issue a little clearer by invoking an
sightful tutorial in the basics of dynamic systems think- example, namely, changing AIDS-risk behavior by
ing in psychology, and in social psychology in particu- means of a television campaign. There exists an exten-
lar. As a commentator, I feel I am more or less in the sive literature on this topic (Fisher & Fisher, 1992).My
position of the interior decorator who is left with the only aim with this example, however, is to illustrate an
assignment of designing the form of the doorknobs or important distinction between time-based and vari-
such highly functional things as the toilet seats, a task able-based models, which leads directly to the heart of
that, in spite of its indisputable intrinsic importance, is dynamic modeling.
not really very attractive. Instead, I have chosen to Let us assume we have a perfect and correct knowl-
readdress two of the fundamental issues raised by Val- edge of the strength of the attitude toward the impor-
lacher and Nowak in order to shed some more light on tance of AIDS prevention in all the members of a
the following questions: What is the fundamental dis- population. Let us also assume that the distribution of
tinction between a dynamic and a nondynamic model these attitude strengths takes the form of a bell curve,
in (social) psychology? and What does dynamic sys- which can be sufficiently specified by its mean MI (this,
tems thinking mean for social psychological theory of course, is an assumption introduced for the sake of
building? simplicity). Next, we introduce an intervention in the
form of an AIDS prevention campaign that consists, for
instance, of the daily broadcast of a television commer-
What Is "Dynamic" in a Dynamic cial that all persons watch with a fixed frequency. Then,
Model? after some time, we apply our absolutely objective
measurement instrument again and find a new attitude
The Longman Webster English College Dictionq strength distribution with mean M2. The whole proce-
defines dynamics as "1. A branch of physics that deals dure can be written down in the form of a model:
with forces and their relation to the motion of bodies.
2. The pattern of change or growth of an object or
phenomenon" (p. 456). Thus, if applied to social psy-
chology, dynamics means something like "the pattern which one should read as: The mean attitude strength
of change of social psychological properties or phe- M2 is the result of applying an intervention I to a
nomena'' or "a branch of social psychology that deals population with mean attitude strength MI. The whole
with variables and their relation to the pattern of change model can be written down in the form of a mathemati-
of social psychological properties or phenomena." With cal identity. Is this a dynamic model? According to the
this definition in mind, it is hard to find something in dictionary definition,it is, because it describes a change

You might also like