Sie sind auf Seite 1von 3

Opinion Article

published: 04 October 2012

doi: 10.3389/fpsyg.2012.00382

Dynamical systems theory in psychology: assistance for the

lay reader is required
Lois A. Gelfand* and Sally Engelhart
Department of Psychology, University of Pennsylvania, Philadelphia, PA, USA
Edited by:
Moon-Ho R. Ho, Nanyang Technological University, Singapore

Dynamical Systems Theory (DST) has gen- in clinical psychology research is that in For example, Hayes etal. (2007) related
erated interest and excitement in psycholog- dynamical models the same variables serve, their study, in which they examined non-
ical research, as demonstrated by the recent in a sense, as both dependent and independ- linear trajectories of depression change dur-
statement, the dynamical perspective ent variables. Another way of saying this is ing treatment, to DST, although there were
has emerged as a primary paradigm for the that dynamical systems are, by definition, no dynamical components to their model.
investigation of psychological processes at feedback models. That is, their dependent and independent
different levels of personal and social real- For example, X(t+1)=aX(t) constitutes variables were distinct (i.e., no feedback),
ity (Vallacher etal., 2010, p. 263). an extremely simple dynamical system with and their data analyses were conducted via
What is less clear to the authors is the one variable (X) and a constant (the coef- static approaches (non-linear growth
degree to which this excitement is justi- ficient a) that, multiplied by X at time t, curve analysis via HLM). Nonetheless, they
fied. Like many psychology researchers, we defines X at time t+1. described the focus of their study using DST
were initially unfamiliar with the concepts, In contrast, models in which dependent terminology (e.g., critical fluctuations, p.
terminology, and techniques used in DST variables are distinct from independent 410). By doing so, they may have led lay
modeling (an approach that was devel- variables, such as OLS regression and hier- readers to conclude, incorrectly, that the tra-
oped in physics), which made it difficult archical linear modeling (HLM, which can jectories of depression change they reported
to judge applications of DST in the articles also be used to perform non-linear growth fit into a DST framework, that their analyses
we encountered. After reading introductory curve analyses), are not feedback models, constituted applications of DST theory and
DST material, we developed some opinions and thus are not dynamical systems. methods to clinical psychology, and that
about how authors of DST-related articles In a 1994 review article, Barton seemed judgments of the presented research would
could help psychologists who are not famil- to blur this distinction, implying that all be relevant to judging the usefulness of DST
iar with DST [hereafter referred to as lay statistical models are dynamical and differ in clinical psychology research.
reader(s)] to begin to make judgments primarily in whether they involve linear or To maintain the distinction between
about their work. In order for DST to be a non-linear equations. dynamical and non-dynamical models,
useful methodology for psychology research, researchers reporting on non-dynamical
From a mathematical perspective, dynam-
we believe, DST-based work must be reason- ics can be thought of as linear or non- models can simply omit any reference to
ably accessible to other psychologists. linear Linear equations are the DST. Researchers presenting non-dynam-
Although we are restricting our discus- cornerstone of statistics. When we per- ical models who choose to refer to DST
sion to the application of DST methodology form an analysis of variance or enter data terminology should explicitly state that
to clinical psychology research, we believe into a multiple regression equation, we their models are not dynamical, and, fur-
that the following three recommendations are using linear equations to describe the thermore, should make clear what the rel-
may be applied to the field of psychology relationships among variables (pp. 56). evance of DST is to the presented research.
more broadly. For example, are DST concepts being pre-
In a response to Barton (1994), sented metaphorically? Do the researchers
Maintain a distinction between Mandel (1995, 107) clarified the dis- speculate that a dynamical process under-
dynamical and non-dynamical tinction between dynamical and static lies their data, but refrain from examining a
models approaches (as well as linear and non- dynamical model? If so, what evidence sup-
A defining feature of a dynamical model linear models),such that OLS regression, ports the speculation, and why is a dynami-
is that the values of the variables in a for example, would be considered linear cal model not investigated? Maintaining a
dynamical system at one time are mod- static and non-linear growth curve analy- clear distinction between dynamical and
eled as functions of those same variables at sis via HLM would be considered non- non-dynamical models will assist the lay
earlier times. One characteristic, therefore, linear static. However, this distinction is reader in making judgments about the
that distinguishes dynamical models from not always clearly maintained in the psy- usefulness of DST in clinical psychology
the statistical models commonly applied chology literature. research. October 2012 | Volume 3 | Article 382 | 1

Gelfand and Engelhart Dynamical systems in psychology

Maintain a clear distinction when in fact they are inconsistent both with data could help readers gain a better sense
between influences on the the specific model and also with the deter- of how the patterns described in other ways
variables from the proposed ministic framework of DST more generally. (e.g., other plots, text descriptions) would
model and other influences. In a different type of example, Chow look in empirical data. Readers might be
DST, by its nature, involves the study of pro- etal. (2005) used a linear dynamical model able to get some sense of the plausibility of
cesses that unfold over time in a determin- to describe periodic fluctuations in hedonic the model based on the look of these simu-
istic manner (absent any perturbations), level. Empirical results showed a weekly lated time series plots. When empirical data
from an initial state, based solely on the periodicity in hedonic level; specifically, the have been collected (e.g., Cook etal., 1995;
functional relationships among the vari- undergraduate subjects in the study, who Gottman etal., 1999; Chow etal., 2005;
ables in the system. In the context of clini- were studied in their natural environment, Boker and Laurenceau, 2006; Fisher etal.,
cal psychology, it may be difficult to identify enjoyed themselves more on weekends than 2011), presenting time series plots of these
variables that operate in such a determin- on weekdays. data along with time series plots generated
istic manner or to construct models that In this case, the hypothesized dynamical from the models (on the same scale), would
adequately characterize their interactions. model posited was a simple model in which allow the reader to get a visual sense of how
Difficulties may arise from a number of hedonic level at one time varied only in rela- well the models fit. If different models are fit
sources, including the intentional actions tion to hedonic level at a previous time. to the same data (e.g., Hufford etal., 2003;
of participants and the difficulties in isolat- However, the fact that the hedonic cycle Witkiewitz etal., 2007), a time series for
ing psychological variables from the myriad appeared to be entrained to a weekly calen- each model should be included. The time
environmental influences that affect human dar cycle suggests that other influences may series plots should be generated at the same
beings. Unless DST researchers explicitly have been at work (i.e., behavioral demands level (e.g., individual, group) for which the
state otherwise, lay readers may assume that are different on weekdays versus weekends dynamical system variables are described in
any influences mentioned by the research- for students). Because Chow etal. (2005) the model.
ers originate from the proposed model, did not explicitly state that the entrain- In our opinion, because time series
and thus be unable to accurately assess ment of periodic hedonic fluctuations to plots do not require technical expertise
the usefulness of the model. Therefore, an external environmental (calendar) cycle for interpretation, showing empirical
when researchers discuss the influences on was not part of their model, the lay reader and modeled time series plots together
variables they examine, we believe that it is may assume, incorrectly, that it was. is a way of presenting results that is par-
incumbent on them to be particularly care- When researchers present DST-related ticularly accessible to lay readers. While
ful in distinguishing between those influ- work, it might be helpful for them to include we did not encounter this kind of pres-
ences arising from the proposed model and two separate sections in the discussion: one entation in any of the articles relevant to
other influences. for influences on the variables that arise clinical psychology that we looked at, an
For example, Peluso etal. (2012, 51), from the proposed model; and another for example from the biological sciences can
presented non-linear dynamical models of other influences. This would put lay readers be found in Figures 4B and 4D of Shiferaw
the changes over time of psychotherapist in a better position to judge the usefulness etal. (2006). In these figures, an empirical
and client emotional valences, in which each of the model, and thus to better evaluate the time series plot of calcium transients in a
participants emotional valence was mod- role of DST in psychology research. stimulated rat cardiac muscle cell is shown
eled as deterministic functions of both the alongside a corresponding time series plot
other participants emotional valence and Include a time series plot from generated from a non-linear dynamical
their own emotional valence at the imme- the EMPIRICAL data or the model, model. Despite slight differences between
diately preceding time. and if both are available, show the empirical and model plots, we believe
The authors also suggested that psycho- them together that the overwhelming resemblance of the
therapists be mindful of and monitor Dynamical systems involve changes in vari- two plots provides a compelling illustra-
these valences and how strongly they impact ables that unfold over time. Although the tion, even to readers with no knowledge of
one another, presumably with the idea that graphical techniques specific to DST are DST or cell biology, of the excellent match
the psychotherapists would adjust their important to include because they show of the dynamical model to the empirical
own values in order to improve therapy specific DST-related properties of models, data. We believe that similar presentations
outcomes. Implicit, then, was an assump- we think that it is also useful to include time in psychology articles would provide much
tion that there were two sources of influence series plots to help the reader conceptualize clearer evidence than model fit statistics,
on psychotherapist emotional valence one the model, to make judgments about the or other statistical measures, of the value
from a dynamical system, in which the emo- plausibility of the model, and, where both of dynamical models.
tional valence changes deterministically, modeled and empirical data are available
and the other from outside of the dynamical and plotted together, to make judgments Conclusion
system, involving direct volitional changes. about how well the model fits the data. Is DST a useful approach for clinical psy-
However, because the authors did not make When results are presented from simula- chology research? Has it already made con-
this distinction explicit, the lay reader may tions in the absence of empirical data (e.g., tributions to the field? We are not sure, and
assume, incorrectly, that volitional changes Peluso etal., 2012), showing a few repre- believe that it is impossible, at this point, for
in variables are consistent with their model, sentative time series plots of the simulated non-expert readers to determine.

Frontiers in Psychology|Quantitative Psychology and Measurement October 2012 | Volume 3 | Article 382 | 2
Gelfand and Engelhart Dynamical systems in psychology

We hope that by providing clearer infor- emotion regulation using a damped oscillator model. linear equations to model the therapeutic relationship.
Emotion 5, 208. Psychother. Res. 22, 4055.
mation about the role of DST models in
Cook, J., Tyson, R., White, J., Rushe, R., Gottman, J., and Shiferaw, Y., Qu, Z., Garfinkel, A., Karma, A., and Weiss, J.
their work, and about the fit of their models Murray, J. (1995). Mathematics of marital conflict: N. (2006). Nonlinear dynamics of paced cardiac cells.
to data, researchers applying DST to psy- qualitative dynamic mathematical modeling of mari- Ann. N. Y. Acad. Sci. 1080, 376394.
chological variables will better enable the tal interaction. J. Fam. Psychol. 9, 110. Vallacher, R. R., Coleman, P. T., Nowak, A., and Bui-
psychology research community to answer Fisher, A. J., Newman, M. G., and Molenaar, P. C. (2011). Wrzosinska, L. (2010). Rethinking intractable conflict:
A quantitative method for the analysis of nomo- the perspective of dynamical systems. Am. Psychol.
these questions.
thetic relationships between idiographic structures: 65, 262.
dynamic patterns create attractor states for sustained Witkiewitz, K., van der Maas, H. L. J., Hufford, M. R., and
Acknowledgments posttreatment change. J. Consult. Clin. Psychol. 79, Marlatt, G. A. (2007). Nonnormality and divergence
We wish to thank Nicholas Forand, Steven 552563. in posttreatment alcohol use: reexamining the pro-
Hollon, Louis Littman, and Robert Rusling Gottman, J., Swanson, C., and Murray, J. (1999). The ject MATCH data another way. J. Abnorm. Psychol.
mathematics of marital conflict: dynamic math- 116 (2): 378.
for their comments and suggestions, and
ematical nonlinear modeling of newlywed marital
Martin Gelfand and Jeff Gerecht for verifi- interaction. J. Fam. Psychol. 13, 3. Received: 23 May 2012; accepted: 14 September 2012; pub-
cations and clarifications regarding dynam- Hayes, A. M., Feldman, G. C., Beevers, C. G., Laurenceau, lished online: 04 October 2012.
ical systems theory. J. P., Cardaciotto, L. A., and Lewis-Smith, J. (2007). Citation: Gelfand LA and Engelhart S (2012) Dynamical
Discontinuities and cognitive changes in an exposure- systems theory in psychology: assistance for the lay
References based cognitive therapy for depression. J. Consult. reader is required. Front. Psychology 3:382. doi:10.3389/
Barton, S. (1994). Chaos, self-organization, and psychol- Clin. Psychol. 75, 409. fpsyg.2012.00382
ogy. Am. Psychol. 49, 5. Hufford, M. R., Witkiewitz, K., Shields, A. L., Kodya, This article was submitted to Frontiers in Quantitative
Boker, S. M., and Laurenceau, J. P. (2006). Dynamical S., and Caruso, J. C. (2003). Relapse as a nonlinear Psychology and Measurement, a specialty of Frontiers in
systems modeling: an application to the regulation dynamic system: application to patients with alcohol Psychology.
of intimacy and disclosure in marriage, in Models use disorders. J. Abnorm. Psychol. 112, 219. Copyright 2012 Gelfand and Engelhart. This is an open-
for Intensive Longitudinal Data, eds T. A. Walls and Mandel, D. R. (1995). Chaos theory, sensitive dependence, access article distributed under the terms of the Creative
J. L. Schafer (New York: Oxford University Press), and the logistic equation. Am. Psychol. 50, 106107. Commons Attribution License, which permits use, distribu-
195218. Peluso, P. R., Liebovitch, L. S., Gottman, J. M., Norman, tion and reproduction in other forums, provided the original
Chow, S. M., Ram, N., Boker, S. M., Fujita, F., and Clore, M. D., and Su, J. (2012). A mathematical model of authors and source are credited and subject to any copyright
G. (2005). Emotion as a thermostat: representing psychotherapy: an investigation using dynamic non- notices concerning any third-party graphics etc. October 2012 | Volume 3 | Article 382 | 3