Are the most creative individuals also the most prolific? Examples of Picasso, Edison, and Freud, to name only a few, would support this idea. On the other hand, Mendel only published eight papers. How does age affect creativity? Goethe was prolific until his death, but Mascagni faded at 26. In short, what accounts for differences in creative productivity? Simonton proposes a model to predict relations between age and creative productivity. The model includes four parameters: 1) initial creative potential, 2) age of career onset, 3) ideation rate, and 4) elaboration rate. To the extent that these parameters produce curves consistent with the data, the mechanisms they represent may have psychological reality.
Selection Variation at Multiple Levels of Analysis
Simonton’s model is based on the general assumption that creativity is a Darwinian process whereby "some selection-variation process (or set of such processes) . . . generates and winnows out numerous conceptual combinations" (p. 67). He makes two assumptions in the model. First, the variation process is blind. That is, an individual cannot predict which combination of ideas will be fruitful. The implication is a random distribution of useful and useless ideas across and between creators. Simonton is quick to point out that this is not the same as saying that the idea generation process cannot be restricted. But, an creative contribution must go beyond what has already been done. For example, in order to qualify as a creative contribution, a research study must ask a question that has not been answered. Thus, "for significant acts of creativity, a point is reached where the creator has minimal guidance from logic or past experience and thus must not rely on an effectively nondirected search for new ideational variations among the population of relevant concepts" (p. 67).
Second, selection-variation processes operate at multiple levels: cognitive, interpersonal, and socio-cultural. At the cognitive level, the selection-variation process operates on the individual’s ideas, which are translated into articles, paintings, poems, etc. At the interpersonal level, it operates on products published by individuals within a domain. Finally, whether a particular contribution will have an impact on other scientists in other disciplines is governed by the selection-variation process. The integration of the assumptions is summarized as follows: "Because the selection processes function at so many different levels, the variation procedure that happens at the cognitive level must be necessarily blind to the ultimate reception of any given conceptual combination" (p. 67).
Simonton begins with the assumption that an individual is capable of a certain number of ideational combinations (N). However, only a fraction of this number will be useful, workable ideas. He defines m (sN where 0 < s < 1) as initial creative potential. Workable ideas are partitioned into 1) amount remaining (x), 2) ideas that need development (y), and 3) ideas translated into completed products (z). Thus:
m = x + y + z
With each additional year of career age, x decreases while y and z increase. The parameters change over the career as follows. First, there is a direct relationship between the rate at which creative potential is used up and amount of creative potential remaining. That is, creative potential is used up faster when more potential remains than when less potential remains. Second, the rate at which products are produced is directly proportional to the number of ideas that need to be elaborated. "For example, the more ideas that exist in the notebooks, the more projects can be going on simultaneously, and the more cross-talk across ideas, thereby stimulating higher publication rate" (p. 68). Finally, the number of ideations is the difference between the "rate at which ideas are being added to the notebooks or sketchbooks and the rate at which ideas are becoming completed contributions . . . " (p. 68). The rest is calculus.
The curve derived from the parameters represents generation of original combination of ideas. This presents a problem because of the impossibility of tabulating the number of original ideas over a career. Creative contributions in the form of publications, songs, symphonies, etc. can be substituted given the assumption that creative productive products are invariant across age with respect to the number of original ideas they reflect. Simonton illustrates: "For example, if researchers are counting scientific journal articles, they assume that each constitutes roughly the same unit of cognitive investment in terms of the number of ideational combinations contained" (p. 69). A weighting scheme can overcome inequivalencies in ideas across products (e.g., song vs. symphony).
An important methodological error in relating creative productivity and age is called the compositional fallacy. This occurs when an aggregated function—a summary curve showing number of contributions at each age collapsed across people—does not describe the relation for any individual and because of differences in life span. That is, a person who is dead cannot contribute. Thus, an aggregated curve might indicate fewer contributions at age 70 simply because individual who could have contributed at that age have already died. A way to avoid this problem is to avoid this problem is to restrict analyses to periods in which all creators were alive, or to individuals who live past a certain age (e.g., octogenarians).
Simonton explains the importance of aggregation curves, assuming that the compositional fallacy has been avoided. Basically, there are aberrations in individual curves that are attributable to any number of factors—e.g., child birth, divorce, illness, etc—that may distort the age-productivity relation and provide an inaccurate picture of the actual state of affairs. Such idiosyncrasies limit the generalizability of findings for one individual to others. Thus, "the researcher must assume that many of these influential conditions are randomly distributed across the life span, and hence averaging across many separate careers will yield a summary curve in which all the random factors have canceled out" (p. 70).
Regression equations with polynomial terms can predict observed data as well as a model like Simonton’s. The advantage of the model is related to the comment in the opening paragraph of this summary about psychological reality. Simonton states:
Only the estimated parameters of the present longitudinal model can claim to have psychological meaning. Unlike the regression coefficients for linear, quadratic, and cubic terms for a third-order polynomial, which lack substantive interpretations, the ideation and elaboration rates have theoretical content. Where a tells us how fast ideational combinations are first emerging during the creative process, b tells us how fast those combinations are then elaborated into presentable ideas.
Parameter estimates are derived with the nature of the domain in mind. They must make sense when compared to parameter estimates from other domains. For example, ideation and elaboration rates in poetry are probably faster than in novel writing. In fact, theoretically generating curves support this idea.
Quantity vs. Quality
What counts as a creative product? Lehman examined age-productivity curves for creations that had an impact. Obviously, the problem is defining impact—the significance of a contribution. Quantity measures (e.g., number of publications) have the advantage of objectivity. But are the same curves found for quantity and quality measures? Simonton argues that quantity and quality are strongly related. This idea is predicated on the blind selection-variation model, but is also empirically supported. For example, the ratio of major works to total output (called the quality ratio) is stable across the career. That is, the relationship between attempts and hits is linear and positive.
Simonton distinguishes among three landmarks: first contribution, best contribution, and last contribution. A contribution is operationalized according to some criterion, such as number of peer citations. The model makes the straightforward prediction that the best contribution should occur at the peak of the age-productivity curve, that is, when the individual is most productive. This is consistent with the finding that across field best contributions tend to vary with curve maxima. Predictions about first and last contributions are less precise. For domains whose age-productivity curves are steeper pre-peak, first contributions will occur earlier, and for domains whose age-productivity curves are steeper post-peak, last contributions will occur earlier. For example, poets will produce their first contribution before historians, but will burn-out faster. Predictions about first and last landmarks are based on the assumption that 1) the quality-quantiity ratio is relatively constant across domains (i.e., the equal-odds rule applies across domains) and 2) the ratio of number of ideas produced to number of products produced is relatively constant across domains.
Research shows that the distribution of lifetime contributions in most fields is extremely positively skewed: individuals in the top ten percent of the distribution account for 50% of contributions. Individuals in the lower half account for only 15%. This finding is predicted by the Price Law, which states that the square root of N, where N is the number of contributors in the field, is the number of individuals who will account for %50 percent of creative contributions. This law applies both to lifetime output estimates and to major works. This again illustrates the point that quantity and quality are highly related. Simonton does acknowledge that the relationship is not perfect (r = .5 - .7), and that some individuals might have higher hit rates (perfectionists, e.g., Mendel?) or lower hit rates (mass-producers, e.g., Picasso?). Nevertheless, in general, "The more total products, the more successful products—and the more unsuccessful products" (p. 77).
Simonton claims that "To predict how productive a creator is going to be in a particular career interval, knowing who the person is is far more useful than knowing how old the person is" (p. 77). This claim is related to the observation that there is substantial variability in age of first, best, and last contributions. The model longitudinal model accounts for these individual differences in the following way. First, individual differences in initial creative potential (m) will account for differences in the height of the curve. Simonton assumes that the distribution of m is predicted by the Price Law—in other words, it’s positively skewed. Second, individuals differ in age of career onset. Simonton argues that these two variables—initial creative potential and age of career onset—should be uncorrelated because they have difference development determinants. Age of career onset is determined by the number of years one has been engaged in acquiring knowledge. Simon, Ericsson and others claim that about 10 years are required to acquire sufficient knowledge for the "combinatory process." Differences in initial creative potential is determined by the richness of the associations between domain-relevant ideas, and this is predicated on "a whole host of genetic, familial, and educational variables" (p. 77).
The contrast of initial creative potential and age of career onset lead to a Creative Potential x Career Onset arrangement. Four predictions emerge: 1) total lifetime output will be negatively related to age of first contribution and positively related to age of last contribution; 2) there is no correlation between age of maximum output and lifetime productivity; 3) age of maximum output correlates positively with age of first and last contribution—the earlier the first contribution, or the earlier the last contribution, the earlier the peak in output; 4) age of best contribution correlates positively with both age of first and last contributions; 5) the correlation between the interval between career onset and first contribution is negatively related to total output.
Longitudinal Stability of Cross-Sectional Variability
Simonton notes that there is considerably evidence to show that cross sectional differences tend to show stability across time: "Those who are the most prolific in the early part of the career are the most productive at the career peak as well as the most productive toward the end of the life" (p. 81). From a behavioral perspective, this is explained as a reinforcement phenomenon in which those most productive in, say, their 20s are reinforced and continue to be productive in their 30s. This model predicts a simplex correlational structure in which the correlations between productivity decrease the farther apart the time intervals. Thus, productivity in the 20s is strongly related to productivity in the 30s, but less strongly correlated with productivity in the 40s.
Simonton’s model makes different predictions. He predicts that the major determinant of creative productivity is initial creative potential. That is, "Individuals with high initial creative potential are the most prolific in the first decade of their career, the most prolific at the peak of their career, and the most prolific in the last decade of their career—and the most productive in all intervening periods as well" (p. 82). This assumption leads to the prediction of invariance in correlations between productivity across time intervals—that is, non-simplex structure. A one-factor model accounts for longitudinal stability in cross-sectional variation. Simonton proposes that this variable is initial creative potential.
Comments and Questions
Simonton’s model accounts for a number of phenomenon: age-productivity curves for individuals, longitudinal stability of cross-sectional variability, and the placement of career landmarks. The model includes two information processing variables—ideation and elaboration rates—and two individual difference variables—age of career onset and initial creative potential and one major assumption—the equal-odds rule, which states that the ratio of total output to creative output is constant. The model is based on the Darwinian idea that a blind selection mechanism operates on new ideas—some are useful and some are useless, but the creator cannot predict.
A number of weaknesses concern mathematical assumptions. The model is also criticized on the grounds that Darwinian models of blind-selection ignore influences of goal-directed behavior. I anticipated another important limitation: "even though a measure of the age at career onset and a record of the domain of creative activity may be obtained, the level of initial creative potential cannot be known in advance" (p. 83). Forecasts can only be made after some amount of time in the career. Simonton argues that "early biographical antecedents of creative potential" might improve the forcast about initial creative potential.
Another problem concerns the nebulousness of the model’s two information processing variables: ideation and elaboration. These variables are not explicitly linked to cognitive mechanisms. Thus, "cognitive psychologists who hope this model helps them pinpoint the precise processes that underlie creativity are going to be severely disappointed" (p. 84). A link needs to be established between computational models of the creative process and models designed to predict creative output.
How and when are parameters estimated in Simonton’s model? He states, "The more simple, abstract, or finite the array of concepts that the creator deals with, the faster the information-processing rates. In contrast, the more complex, associatively rich, or unbounded this repertoire, the slower become the same rates" (p. 233). The nature of the cognition involved in the activity influence parameter estimates.
An important variable in the model is age of career onset. Individuals who begin their career earlier will show peak productivity earlier, and first and last career landmarks will occur earlier. However, initial creative potential and age of career onset are unrelated. What factors influence age of career onset? Extraneous (e.g., social) factors may have an influence. The availability of role models may influence individuals to start the knowledge acquisition process that is necessary for creative output earlier. (Recall that initial creative potential is predicated on 1) amount of domain-relevant knowledge and 2) richness of knowledge associations, which may be in part genetically influenced.) What factors influence the choice of a domain is also an interesting question. Personality traits, which have a heritable component, may have an influence on this decision.
What factors influence level of initial creative potential? Simonton argues the interaction of genetic and environmental factors. Intelligence, he claims, is one determinant that has a genetic component. The manner in which information is organized is also important. He states, "the cognitive material must be profusely interconnected to nourish the free-associative variation process that generates creative ideas" (p. 245). He links such an organization to genetic endowment. He states, "I am not claiming here that creators are crazy. On the contrary, the research suggests that these individuals lie on the borderline between normal and abnormal thoughts" (p. 245).
Comments and Questions
Simonton’s model can be applied to individual as well as to aggregated data. It shows impressive predictive power for individuals. To review, the model includes two individual difference variables: 1) age of career onset and 2) initial creative potential. The former variable is known a priori; the latter variable is generated by the model. (Two other variables—ideation rate and elaboration rate--are fixed for a particular domain, but vary across domains.) Would it be possible to obtain estimates of initial creative potential for individuals and then to correlate this variable with other variables. Of course, the complication is that age-productivity curves are estimated for individuals who are no longer living. However, it seems that it would be possible to obtain estimates for people who are living—perhaps for octogenarians—and then to investigate correlations of this variable with other variables (e.g., intelligence, personality traits, environmental variables, etc.). It might even be possible to examine relations based on biographical information, although this would be difficult. An important question is: What is the distribution of m? If the distribution is not normal, then some transformation would be required. Simonton assumes that "the amount of material available for free variation is normally distributed in the population of creators in an enterprise" (p. 76). But this is not the same quantity as initial creative potential—that is, the useful proportion of ideation combinations (m=sN). The distribution of N (number of ideational combinations) is lognormal (positively skewed), so I assume that the distribution of m would also be skewed.