MIND, CULTURE, AND ACTIVITY, 4(1), 34-41 Copyright© 1997, Regents ofthe University of Califomia on behalf of the Laboratory of Comparative Human Cognition Chaos, Competition, and the Necessity to Create David K. Dirlam King College In a study of the frequeticy of usage of developing strategies in 5-year-old to 18-year-old children's drawings and in 1968-1992 developmental research strategies, Dirlam (1996) showed that iti both cases, usage declined exponentially as successor strategies emerged. This "general gamma" model revealed that macrodevelopment involves (a) inevitable declines in primitive strategy usage, (b) simultaneous use of multiple strategies, and (c) cultural mediation of strategy change. Two basic questions remained. Why does macrodevelopment involve decline, and what happens after the most advanced strategy declines? A new chaos-theory model, rooted in population growth models, assumed that the rate of strategy decline was related to the summed competition from growing, successor strategies. It fit the data as well as the general gamma and revealed that if no successor to an advanced strategy occurs, its usage usually becomes chaotic. This implies that developing systems of strategies cannot maintain stability without creatively introducing new strategies. Dirlam (1996) found that frequencies of use over time of multidimensional strategies of both children's drawings and developmental research obeyed the general gamma probability law; that is, each strategy declines following a negative exponential curve as it is replaced by its successor. This implied that both were mediated by the same process. The developmental research strategies, most of which were defined from Danziger (1990), were found in a random selection of 599 articles from Child Development and Developmental Psychology, 1968 to 1992, whereas the drawing strategies, defined from Lowenfeld (1957) and Piaget and Inhelder (1967), were found in drawings obtained from 1,222 5-year-old to 18-year-old children in northeastern New York. The general gamma fits were precise enough to allow accurate predictions—for example, that the latest developmental research strategy (analytical, contextualized research) would reach peak usage between the years 2004 and 2010 and decline thereafter. This strategy uses both difference and correlational statistics, multiple social contexts, and well-described environmental settings. A change in any one of the three dimensions ofthe strategy would compose a new strategy. Three processes of strategy development were distinguished by the manner in which the strategies were preserved: evolutionary strategies by genetic transmission, developmental strategies by automatization, and historical strategies by cultural transmission. Because developmental research methods could not possibly arise from a genetic basis or from a solitary individual's interaction with the environment, the common process was considered to be cultural. Thus, a particular drawing strategy used by a child was considered to be a cultural artifact, as in Cole's Requests for reprints should be sent to David K. Dirlam, Department of Psychology, King College, Bristol, TN 37620-2269. E-mail: dkdirlam@king.bristol.tn CHAOS, COMPETITION, AND NECESSITY TO CREATE 35 (1995) conception, which is mediated by a cultural agent, such as an admired peer. This is much in the same way as the mediation of English children's games studied by the Opies (1969) or of the literary genres discussed by Cooper (in press). Two basic questions arosefromthe general gamma model. Why does macrodevelopment involve decline? What happens after the most advanced strategy declines? Progress on these questions arose from a suggestion by Danziger (personal commtmication, July 1996) stemming from the issue of historical prediction. He suggested that I take another look at chaos-theory models for the historical data. Two previous attempts (in 1988 and 1992) failed to tum up a suitable model. This time, however, instead of focusing on the equations, themselves, I created spreadsheet examples. In his excellent introduction to the use of chaos models in developmental theory. Van Geert (1994) defended spreadsheets for being as practical for building developmental models as concrete is for building warehouses. I agree, and I add that they are peculiarly well-adapted to the task. The use of spreadsheets requires that models be developed in a manner remarkably similar to the developmental process itself. First, a formula is entered into a particular cell. This produces an outcome that is displayed in that cell. Then, the formula is copied to other cells, where it is automatically adjusted so that each new cell operates on the results from the prior cell. Textbooks on chaos theory often give the impression of acomplex branch of differential equations with strange visual realizations. A spreadsheet, on the other hand, reveals simple, iterated algebraic formulae. Iteration in each new cell builds on the outcome in its prior cell, much like scientific projects build on the outcomes of preceding projects. Besides the naturalness of spreadsheet iterations, graphical displays can be added to the same sheet as the formula. This makes the computational results immediate, like having a micro/macroscope that allows one to view a new generation of cells and the new flower they comprise simultaneously. In addition to intriguing stirprises in the flowery graphs, I found a model that fit my data as well as the general gamma, but it did not so much replace that older model as explain it. EVOLUTIONARY STRATEGIES AND POPULATION GROWTH In 1798, Malthus (1798/1959) proposed that populations grow exponentially, like compound interest, the number in each new generation being l-t-r times the number in the last generation. This relation can be expressed in the simple formula (see Figure 1): x'=[l+r]x (1) In this formula, each new value of jc' depends on the last value of jc. Before the advent of computer programming, a little algebra would be used to convert this formula into one with a single unknown, producing the familiar, "exponential" growth equation used for compound interest. Because spreadsheets make the conversion no longer necessary, we can leave the equation in the simpler form. In 1846, P.-F. Verhulst proposed that population growth does not proceed indefinitely, but that the rate of growth declines as the population approaches a maximum sustainable value. If A: is the proportion of this maximum, then he suggested that the rate is reduced by the amount of possible growth remaining. In other words, the rate becomes r(l—x). This produced his now famous "logistic" model of population growth (see Peitgen & Richter, 1986; the underlined portion is what Verhulst added to Malthus's equation): 36 DIRLAM 2S0O . 2000 . / 1S00 , / 1000 . 500 . 0. 1 1 1 •"1 xO 1 1 25 20 ^ 3D 1 35 Generations FIGURE 1 Malthusian (exponential) growth 20% growth rate from two individuals. A :A A f U A h A^iA A A A A • 10 15 • 20 : . > : : 25 '• ' 30 • : : ; 35 40 Generations FIGURE 2 Proportion of maximtim population produced by selected rates of Verhulstian growth. x'=\l+r(l-x)]x (2) With common growth rates, the new term r(l-x) causes a decline in the growth as x approaches 1 (the maximum sustainable population; see the line of wide dashes in Figure 2). Notice that with r < 0, there is exponential decay (the wide dotted line in the figure). This fact combined with the evolutionary implication of Verhulst's equation suggested its usefulness for explaining why development might proceed via decay. It also suggested some intuitively satisfying features of development that are not readily apparent in the general gamma. With changing positive values of r, X approaches an equilibrium (r < 2, the wide solid line of Figure 2), oscillates (r > 2; the thin black line), becomes chaotic (r > 2.57; the thin dotted line), and takes on nonsensical negative values (r > 3; not shown). Exponential decay, sigmoid growth, oscillation, and chaotic fluctuation are all common developmental pattems. CHAOS, COMPETITION, AND NECESSITY TO CREATE 37 LOGISTIC MODELS OF COMPETING STRATEGIES Strategies do not multiply in a vacuum, but compete at every opportunity with other populations, or "growers" as they are currently referred to in the developmental literature. Van Geert (1991) introduced the idea of two competing growers, each with different environmental carrying capacities, k, and affected by both the level of a second grower, y, and its competitive strength, c: x' = [l-\-r(l - xlk) - cy\x. Clayton and Frey (in press) modified Van Geert's equation to model competition between two growers that share the same capacity: x' = [l+r(l -(x + y) / k}- cy]x. Setting the carrying capacity to 1 had no adverse effect on the model's fit and reduced the number of pariimeters. Note that the rate, r, is now reduced by x -h y (not just x alone). Also, note that cy is subtracted from the total amount of increase of each generation. This allows for a reversal of growth (i.e., decay), when the competition is high. The result with the underlined portions modified from Verhulst's equation is the following: x' = [l+r(l - (;c + -v)Vcv1;c (3) I generalized the Clayton-Frey model to any system of growing sfrategies in which competition is directional, with later strategies suppressing earlier ones but not the reverse (j refers to strategies with later mean times of appearance than the one in question and i refers to all strategies; as before, the underlined portions were modified from the prior numbered equation): jc'=[l+Kl-&)-£ca]j«: (4) Equation (4) needs to be restricted so thatx' never goes below zero (in spreadsheets, this requires inserting the function that chooses a maximum value of 0 or the formula outcome). When each JC' has been calculated, the probabilities can be determined from the use of all strategies in the system (i.e., the maximum possible usage): P(x^-xiljc'i (5) Finding the least-squares differences between (5) and the raw data resulted in fits to both the drawing and research strategies data that were as good as the generalized gamma distributions (see Figures 3 and 4). The complexity ofthe two models is also comparable (cf. Myung & Pitt, 1995): (a) botli require a rate parameter; (b) adjusting the starting values {xo) for the competing strategies system is comparable to combining cells to get frequencies greater than 5 for the general gamma; and (c) selecting competition weights is comparable in complexity, but more flexible than selecting predecessors. The excellentfits reinforce the discussion in Dirlam (1996) conceming the similarity of drawing and research sfrategy development. They suggest in addition that the reason why development proceeds via decay is that later strategies cause the use of earlier strategies to fade. THE SURVIVAL VALUE OF COMPETITION AND THE PERSONAL AND SOCIAL NECESSITY OF CREATIVITY The stability of all but very small competing strategies systems depends on perpetually adding new strategies. By the time several strategies have been replaced, new strategies need to have such a. a •O '. I .5 •p .4 •T- • \ *^ A / \ . 1 .2 y •• - X /N. p. . 1 /i Q .' -p— - 'iS 'JC&^ 1 i' • -ffTir-lf-^ • W*^ • * • 0.5 4.5 1.5 2.5 3.5 .1 5.5 6.5 7.5 8.5 Years after age 5.5 9.5 - - -1- - -11 no shape control-no organization - - O - -12 no shape control-baseline - - A - -13 no shape control-overlap - - • - - 22 line control-base line — O — 3 2 curve or soiid-i^aseline - - A^ - -23 line control-overlap - - • - - 24 line control-plane or cameo — A — 3 3 curve or solid-overlap — H — 3 5 curve or solid-plane — 0 — 3 4 cunre or sdid-cameo — A — 4 3 curve and solid-overlap —X—..Designs — • — 4 4 curre and solid-cameo — • — 45 curve and solid-plane 10.5 11.5 12.5 13.5 b. - - -1- - -11 no shape control-no organization - - O - -12 no shape control-baseline - - A - -13 no shape control-overlap - - • • • - -22 line control-base line ——O 32 curve or solid-baseline - - * - - 2 3 line control-overlap - - • - - 24 line control-plane or cameo — f t — 3 3 curve or solid-overlap — B — 3 5 curve or solid-plane • .5 .4- '• _ 3 ' I ) $ , ' ' Q • . - • • . .2 .1 *' 0.5 1.5 — A — 43 curve and solid-overlap —X—..Designs — • — 4 4 curve and solid-cameo — • — 4 5 curve and solid-plane ^ ' 2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 13.5 Years after age 5.5 C. Q 1 .5 .4 3, • * " • .2 — O ^ — 3 4 curve or solid-cameo — A — 43 curve and solid-overlap —X—..Designs — - • — 44 curve and solid-cameo • — • — 4 5 curve and solid-plane .1 * i 0.5 - - - ( - - - ) 1 no shape control-no organization - - O - -12 no shape control-baseline - - A - -13 no shape control-overlap - - • - - 22 line control-base line — 0 — 3 2 curve or solid-baseline - - A - -23 line control-overlap - - • - - 24 line control-plane or cameo — 0 — 3 3 curve or solld-overtap — B — 3 5 curve or solld-plahe 1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5 Yeans after age 5.5 9.5 10.5 11.5 12.5 13.5 FIGURE 3 (a) Raw probabilities of drawing strategy usage, (b) Probabilities of strategy usage according to the general gamma, (c) Probability distributions of competing strategies in drawing. 38 - - A - - 1 1 1 simplification - - O - - 1 2 1 social - - • * • - - 1 1 2 localized — A — 2 1 2 anaiyticai iocaiized --•--122situated 211anaiyUcai 221 anatytical social 222 anaiyticai situated 0.0 12 16 20 Yrs after 1968 - - A - - 1 1 1 simplification - - O - - 1 2 1 social --*--112locaii2ed — A — 212 anaiyticai Iocaiized - - • - - 1 2 2 situated — D — 2 1 1 anaiyticai — A — 221 anaiyticai social — • — 2 2 2 anaiyticai situated 12 16 years after 1968 -A--111sitnptiflcation - O - - 1 2 1 sociai - - * - - 1 1 2 iocaiized A— 212 anaiyticai iocaiized -•--122situated 211 anaiyticai 0.1 221 anaiyticai sociai 222 anaiyticai situated 12 16 Yrs after 1968 FIGURE 4 (a) Raw probabilities of research strategies, (b) General gamma probabilities of usage, (c) Competing strategies system for research. 39 40 DIRLAM high growth rates to get established that in the absence of competition from later strategies (i.e., in the absence of innovations), their continued growth will become chaotic. If the strategies in question are evolutionary, then the low points in chaotic growth (see the thin dotted line in Figure 2) make a grower particularly vulnerable to extinction. This finding implies that the existence of an effective competitor increases a species' chance of survival. In the context of developmental strategies, as any nondrinking alcoholic is aware (see Lave & Wenger, 1991), the extinction argument does not apply. Developmental strategies may return after falling to zero frequency for long periods of time. Likewise, apparently abandoned historical strategies (e.g., genocide or even human sacrifice) may resurface with agonizing frequency. Even so, Clayton and Frey (in press) noted that despite their high growth rates, chaotic growers do not compete successfully against slower rate, nonchaotic growers. These conclusions strongly support the value of innovation to both individual and social stability. So what happens to the prediction of the demise of the analytical, contextualized approach in developmental research? The competing strategies model suggests that it is possible for this last strategy not to decay in the time frame predicted by Dirlam (1996), because no competitor may be found. But if a successor strategy is not found, then we can expect the last strategy's usage to begin fluctuating in a chaotic fashion, endangering the whole enterprise. A reasonable and testable projection is that within a decade, chaos modeling will have sufficient competitive strength to begin causing the usage of difference combined with correlation statistics to level off. To be a competing developmental research strategy, and not just a new analytical technique, however, such modeling will need to be combined with the multiple social contexts and well-described environmental settings that characterize sociocultural research today. ACKNOWLEDGMENTS Research funds were provided from the Mellon Foundation and the Appalachian College Association. Special thanks go to Kurt Danziger for advice and encouragement through several years ofthe history project, Michael Cole for encouraging this article and its predecessor, Daniel Bowell for assistance in securing sources, Daniel Fetters for clarifying various forms of Verhulst's equation, and Keith Clayton and Barbara Frey for the manuscript that included their competing strategies equation. Step-by-step data analysis details and an Excel example can be obtained by letter or e-mail (dkdirlam @king.bristol.tn.us). REFERENCES Clayton, K., & Frey, B. (in press). Inter- and intra-trial dynamics in memory and choice. In W, Sulis & A. Combs (Eds.), Nonlinear dynamics in human behavior. Singapore: World Scientific. Cole, M, (1995). Culture and cognitive development: From cross-cultural research to creating systems of cultural mediation. Culture and Psychology, 1, 25-54. Cooper, C. R. (in press). Genre and genres. In C. R. Cooper & L. Odell (Eds.), Evaluating writing: Second edition. Urbana: National Council of Teachers of Enghsh. Danziger, K. (1990). Constructing the subject: Historical origins of psychological research. Cambridge, England: Cambridge University Press. CHAOS, COMPETITION, AND NECESSITY TO CREATE 41 Dirlam, D, K. (1996). Macrodevelopmental analysis: From open fields to culture via genres of art and developmental research. Mind, Culture, and Activity, 3, 270-289. Lave, J. R., & Wenger, E. (1991). Situated learning: Legitimate peripheral participation. Cambridge, England: Cambridge University Press. Lowenfeld, V. (1957). Creative and mental growth: Third edition. New York: Macmillan. Malthus, T. R. (1959). Population: The first essay. Ann Arbor: The University of Michigan Press. (Original work published 1798) Myung, J., & Pitt, M. A. (1995). Applying Occam's razor in modeling cognition. Unpublished manuscript. Opie, I., & Opie, P, (1969). Children's games in the street and playground. Oxford, England: Clarendon. Peitgen, H., & Richter, P. H. (1986). The beauty of fractals. New York: Springer-Verlag. Piaget, J., & Inhelder, B. (1967). The child's conception of space. New York: Norton. Van Geesrt, P. (1991). A dynamic systems model of cognitive and language growth. Psychological Review, 98, 3-53. Van Geert, P. (1994). Dynamic systems of development: Change between complexity and chaos. New York: HarvesterWheatsheaf. Verhulst, P.-F. (1846). Deuxieme memoire sur la loi d'accroissement de la population [Second paper on the law of population growth]. Memoires de I'Academie royaie de Bruxelles, 20.