J.L.Lemke Online Office

Review for Theory & Psychology 6(2): 352-356, 1996.


City University of New York
Brooklyn College School of Education
Brooklyn, New York 11210 USA

A Review of:
Linda B. Smith & Esther Thelen, Eds. A Dynamic Systems Approach to Development. Cambridge, MA: MIT Press. 1993. (xviii + 414 pp.) ISBN: 0-262-19333-7.

At first glance, this formidably technical-looking volume seems the sort of book to be archived in the library and referred to only occasionally by a few specialists. But consider the basic questions at the heart of its inquiry: How can we understand and model discontinuous changes in behavior patterns during development? If neither DNA nor neurophysiology operate in ways that can predetermine development, how does development result in the orderliness we construe in behavior? If the information that corresponds to this orderliness is not put in at the beginning, how can it arise during development? And if the information is not intrinsically internal, how should we redefine the system in which it does arise, the system that develops, that learns, that acts? These are not merely technical questions of some esoteric branch of developmental psychology or mathematical modelling theory. They are problems that challenge the ruling paradigms of today's perceptual, motor, and cognitive psychologies.

The 22 contributors to this volume collectively offer us a detailed view of an emerging alternative paradigm. It unifies cognition with its developmental and evolutionary roots in perceptual-motor behavior, shows how development constructs behavioral schemas as its products rather than being guided by them as its sources, and implies that the fundamental unit of analysis in the psychology of behavior cannot be either the human individual or any subsystem of that individual, but but must rather be the complexes of dynamic interactions between individuals and environments, and among processes and subsystems. Probably these researchers would not dare to so radically challenge the dominant "preformationist" paradigms of the field, including Chomsky's innatism, Fodor's modularism, and the very notion that knowledge structures or schemas of competence guide performances and account for their invariant features, if they were not backed by the successes of the new dynamical analysis paradigms in physics, chemistry, and both organismic and ecological biology. Readers of this review will most likely recognize the new paradigm as that of "complexity theory" or "self-organization theory", the mature form of our new insights into the order that lies behind apparent "chaos" in complex dynamical systems. While it remains to be demonstrated just how usefully this paradigm can be transposed into the study of complex human behavior, the contributors to this volume certainly make a good case for at least its heuristic value. They demonstrate new methods of data analysis with interesting findings; they use it to formulate new kinds of experimental and observational questions; and most importantly, they use it to both critique and offer specific alternatives to extremely fundamental theoretical conceptions.

The first subfield to make use of the new mathematical analysis was developmental studies of motor co-ordination (following similar work on adult patterns), and the technical studies in Part I are drawn mainly from this field. Here the type of data available (time-series data on continuously varying behavior described in ordinary space and time) was already well-suited to the mathematical model. This basic model is sketched out in the first chapter, by Kelso, Ding, and Schoner. In it the behavior of a system through time is represented as the movement of a point in a high-dimensional space; this hypercurve is then "projected" onto a much lower-dimensional space of hypothesized critical parameters, where, if we are lucky, we discover miraculously simple patterns lying behind apparent complexity. This "order" results emergently from the couplings and interactions of many separate units in many different ways across many scales of the parameters. In complex systems emergent order seems to be almost inevitable, given a sufficient quantitative degree of interdependence among the processes or units that give rise to it. (One should note that this "orderliness" also in some sense arises from our analytical inclination to project the data in such a way as to construct it. There are always far more possible viewpoints which do not show any humanly or culturally recognizable "order".)

The four empirical studies in Part I illustrate not only the general heuristic uses of the model, but also some of its basic theoretical moves. These are summarized in the useful overview by the editors, and most fully exemplified in the analysis of spontaneous cyclical movements in an infant (Robertson, Cohen, & Meyer-Kress). Order, or coordination, is seen as an emergent property resulting from the more elementary dynamics of constituent processes or subsystems without the need for any localizable source of this order either inside the system or outside it. Indeed some of the studies are at pains to point out that, for example proposed central nervous system sources or governors, could not provide the kind of direction or regulation needed. Characteristically, the studies also find it necessary to redefine the system itself so as to include all the elements which contribute to the emergent pattern of orderliness.

In the simple case of learning to walk, for example, the system has to include the properties of the ground or surface on which we walk. Developmental processes importantly include explorations of possible dynamical strategies for the combined system of child-and-ground, or more generally of child-and-environment. It is in this larger system, best regarded as a system of interdependent processes (i.e. of actions rather than of actors, animate and/or inanimate), that order emerges, and not in the system of the child as isolated entity.

As we move on into Part II, the paradigm is extended, quite naturally from motor-coordination to perceptual-motor processes, generally within the theoretical framework pioneered by J.J. and E.J. Gibson, in which all perception is the result of action, a motor-driven construction of the perceptual interactional world. But the most striking aim of Part II is to carry the program further into the domain of complex cognition and cognitive development.

Smith and Thelen in their long and theoretically ambitious introduction to Part II (a distillation of their companion volume; see below) offer a pounding six-point critique of what they call "structural" accounts of the orderliness of performance (i.e. of constructable invariances from one performance to another). These accounts, they argue, by dividing competence from performance, are led to set up cognitive schemas or structures, whether for perceptual-motor routines or for cognitive operations, as the "executives", the directors or governors of situation-variable performances, which provide the informational source for their invariant features. Smith and Thelen, the other contributors, and indeed the entire self-organization paradigm, argue that this is both unnecessary and, finally, circular. The invariants are emergent; they neither have nor require such sources or executives. The proposed cognitive schema executives are simply the observed invariants restated as causes , when in fact they are themselves rather the products of what is going on in each performance.

Once you make the jump to this new way of looking at the issue, it is hard not to suspect a deep cultural bias in our traditional intuitions about the origins of order. From our computational paradigms of executive control over processes, to the now-repudiated view of DNA as an aloof dictator of cellular biochemistry, or the failing models of centralized CNS control of behavior, our sciences of order have assumed the necessity of there being some localizable seat or source of that order: a patriarch, an elite, a governor, a dictator, a central government, senior management -- some authority with the power to keep the unruly masses of individual instances within some proper bounds. Scientific paradigms so constructed see variation as noise in the system, to be eliminated that we can better understand the invariants, the bounds determined by the ruling element, which is what is "important" about the system or the data. But self-organization theory is more genuinely democratic in recognizing that order can emerge from below, that indeed it almost inevitably will emerge from below, without imposition from above. Variability from case to case is not to be erased as noise, rather it provides critical information, when viewed longitudinally case-by-case, about the different possible pathways to order.

Another modern European cultural bias which may be highlighted by the new paradigm is the naturalization of the human individual as the relevant unit of analysis for development, or for cognition. When the focus is on the relevant processes which contribute to the emergence of order in behavior, those processes most often turn out (as with the role of the ground in learning to walk) to be interactive ones across the boundary of what we traditionally define as the human individual. This is not news, but it acquires a radically new significance when it means that the constituent processes over which cognition itself is distributed (rather than localized) extend beyond the systems of the isolated (and if isolated, soon dead) human body. This paradox was repeatedly put to productive use by Gregory Bateson (e.g. 1972), and is not so far removed the perspectives which gave rise to J.J. Gibson's affordances, Jakob von Uexkill's Umwelt, or L.S. Vygotsky's, A.R. Luria's, and A.N. Leontiev's versions of transmediation. This shift in perspective makes such notions as "situated cognition" or "social cognition" less paradoxical: the system in which cognitive or semiotic activity is taking place is larger than that of individual human organisms. We no longer have the option of localizing a source for the order of behavior at all, much less in one part of the total system of relevance.

Just as studies in Part I included the ground as part of the system in which walking develops, several in Part II examine the development of co-ordinated social activities in which the minimal analytic unit for the emergence of this order is the dyad or group. Why should we suppose that the psychological, or semiotic (meaning-making), unit of analysis should exactly coincide with the biological one? (Even the latter is increasingly trans-organismic.) Or that a persona, which is as much social as psychic, and is defined and constructed by very different discourses and practices and interactions, should map one-to-one onto the same system of relevance as our notions of a biological organism? But such a conflation is exactly what our modern European cultural history has naturalized for us. (For some counter-examples to one-on-one mapping, and consequences for theory, see Lemke, in press, Chapter 5).

These are all themes which developmental psychologists may already find familiar (many exist in Piaget's work, for example, especially in the later work influenced by modern developmental biology, e.g. in Biology and Knowledge, and in the debate with Chomsky, in Piatelli-Palmerini 1980), but which are less easily assimilated by cognitivist paradigms which descend from Herbert Simon's management science and the cybernetic-regulatory models at the root of computer hardware and software design (before parallelism and neural nets). New paradigms call for new metaphors, new analogies. Another of the heuristic values of the self-organization model is that it defines a class of systems to which it applies that cuts across traditional disciplinary boundaries. Ecology, a science already relatively well-along in developing models of self-organization, can now provide possible approaches to the study of cognitive development (as in the chapter by van Geert in this volume). I have also found these metaphors heuristically useful in analyzing human semiotic ecologies (Lemke 1993 and in press, Chapter 6).

One domain of the modern theory of order which is certainly relevant to the issues raised in the book, but which seems rather neglected is that of post-Darwinian evolutionary theory (e.g. Kauffman 1993, Salthe 1993), where the picture is less one of random variation selected by the dictates of the environment, and more one of self-organization in larger systems: co-evolving ecologies of species, and more generally co-evolving ecologies of intra- and inter-organismic processes. The unit of evolution in such models is not even the system-type itself, but the developmental trajectory of typical instances of the type. In this picture we no longer have the option of exporting the source of order to some memory of selections past, embodied in the individual. Evolutionary memory is dynamic: the whole distributed system of cumulative consequences of past selections must recreate the latest configuration (by recapitulation or otherwise) anew each time. What we think we know today about neurological development (e.g. Edelman 1992), if assimilated to this paradigm, also makes it pretty well impossible that brains inherit intact governors of behavior, or even that the detailed neurological participation of different brains in ordering similar behaviors in different individuals (or in the same individual in radically different contexts) is the same.

All this rather pulls the presumptive biological foundations out from under even the strongest cases for innate controllers of cognitive functioning (e.g. Chomsky's innatist interpretation of the invariant syntactical features of language behavior). The chapter on language acquisition in this volume (Tucker & Hirsh-Pasek) tries to offer a complex linguistic model of how, with sufficient (and attested) interdependence among semantic-pragmatic, syntactic, and phonological information in situated language-use, it might be possible for the child to decode the language system without innate guidance. These are valuable suggestions, but they are still couched well within a meta-theory of language that grew up mostly under the old regime. More consistently with the new paradigm we would not see children as little linguists deciphering a code, but as integral constituents of a community, which becomes the system of relevance for language-acquisition, not the child as such or any processes localizable in the child. And this "code" would not be simply language as linguists, with a disciplinary bias towards autonomy, have lately defined it. The order in language-using behavior should emerge from interdependencies within a larger complex of perceptual-motor and embodied cognitive-semiotic processes of the child's participation in the community (as culture and as material ecology).

This complex includes far more than what is isolated from it analytically as language: it includes the sensorimotor processes of aural perception and vocal articulation (with all their wider couplings to other body systems); the original developmental unity of speaking, gesture, and larger body-movement patterns (cf. interactional synchronies); the similar unity of gesture and drawing, of made and seen visual-semiotic patterns; indeed all the forms of information in embodied, interactional activity. Surely in all this, taken together, as it is lived, there is more than enough informational redundancy to account for the self-organization of the developing organism into all the semiotic orders of the community, simultaneously and in concert. The information we need to understand and use language is never all coded in language, at any stage of development.

I hope I have not abused the length assigned for this review in vain, and that when your library buys this book, you will be first in line to struggle with its complexities and its challenges.


Bateson, G. 1972. Steps to an ecology of mind. New York: Ballantine Books.

Edelman, G. 1992. Bright air, brilliant fire. New York: Basic Books.

Kauffman, S. 1993. Origins of order: Self-organization and selection in evolution. London: Oxford University Press.

Lemke, J.L. 1993. Discourse, dynamics, and social change. Cultural Dynamics 6(1): 243-275.

-- . In press. Textual politics: Discourse and social dynamics. London: Taylor & Francis.

Piatelli-Palmerini, M., Ed. 1980. Language and learning: The debate between Jean Piaget and Noam Chomsky. Cambridge, MA: MIT Press.

Salthe, S. 1993. Development and evolution: Complexity and change in biology. Cambridge, MA: MIT Press.