simonton (2007) - creative life cycles

7
Creative Life Cycles in Literature: Poets Versus Novelists or Conceptualists Versus Experimentalists? Dean Keith Simonton University of California, Davis The economist Galenson (2005) proposed a theory of creative life cycles that distinguishes between early-peaking conceptual creators (finders) and late-peaking experimental creators (seekers). This con- trast is claimed to invalidate previous research findings that poets tend to peak earlier than novelists. However, a multiple regression analysis of his published data on 23 creative writers shows that the poetry-novel genre contrast makes a contribution to the prediction of the career trajectory that is orthogonal to the conceptual-experimental contrast. The result is a fourfold typology of creative life cycles: conceptual poets, conceptual novelists, experimental poets, and experimental novelists who do their best work at ages 28, 34, 38, and 44, respectively. The article closes with a discussion of additional empirical and theoretical issues. Keywords: creativity, age, life cycles, literary genres, conceptual versus experimental creators Beyond doubt, the oldest topic in the empirical study of cre- ativity is the relation between age and creative output (Simonton, 1999a). The first scientific analysis regarding this question was conducted by Que ´telet’s (1835/1968) concerning longitudinal changes in the dramatic production of eminent British and French playwrights. About a century later the creative life cycle was taken up by Lehman in a series of articles that culminated in the 1953 classic Age and Achievement. More recent research on this subject has been conducted by numerous other social and behavioral scientists (e.g., Cole, 1979; Dennis, 1966; Over, 1989; Stern, 1978; Simonton, 1991a; Zickar & Slaughter, 1999). In addition, it has been the subject of comprehensive literature reviews (e.g., Lindauer, 2003; Simonton, 1988a, 1996). Not surprisingly, given the sheer mass of research, many key results have been replicated so many times as to constitute among the most secure articles of knowledge in the psychological sciences. Two findings are especially robust (Lehman, 1953; Dennis, 1966; Simonton, 1988a, 1997). First, the output of creative products tends to change over time, rising relatively quickly to a career maximum and then declining somewhat gradually thereafter. Typically, the peak occurs some- time in the late 30s or early 40s, and the productivity toward the end of the career is about half that at the career maximum. This longitudinal trend is approximated by an “inverted backward-J” function (Simonton, 1988a). This function is specified by a second-order polynomial in age where the linear term is positive and the quadratic term is negative. Second, the specific shape of this single-peak function varies according to the domain of creative achievement. For instance, the optimal age for writing poetry tends to be somewhat younger than that for writing novels (Dennis, 1966; Lehman, 1953; Simonton, 1975). Indeed, Simonton (1975) found that the age gap between poets and novelists was invariant across history (from ancient to modern times) and diverse literary traditions (Japanese, Chinese, Islamic, European, etc.). This effect is large enough to help explain why eminent poets tend to have a shorter life span than do novelists (Kaufman, 2003; Simonton, 1974; cf. Kaun, 1991). Be- cause poets produce their best work at a younger age, they can die younger without that event imposing a severe cost on their repu- tation (cf. McCann, 2001). A similar effect holds for mathemati- cians who display both accelerated career trajectories and shorter life expectancies (Simonton, 1991a). The age-output results have been the subject of several theoret- ical interpretations (e.g., Beard, 1874; Diamond, 1984; Simonton, 1997; Stephan & Levin, 1992). Some of these explanations are sociological or economic, whereas others are clearly psychological in nature. In the latter category is a two-step combinatorial model of the creative process that has been developed in a series of articles (Simonton, 1984, 1989, 1991a, 1997) and books (Simon- ton, 1988b, 2004a). In simple terms, this mathematical model assumes that creators launch their careers with a certain amount of creative potential defined as a set of ideas or mental elements available for free combination. This potential then undergoes the two-step process of ideation, by which useful combinations are generated, and elaboration, by which those combinations are con- verted into overt products. This two-step cognitive procedure then yields a double-exponential age function that accurately predicts the career trajectories indicated in previously published data sets (Simonton, 1984, 1989, 1997). Furthermore, because the ideation and elaboration rates are assumed to be domain-specific, the model provides a theoretical explanation for contrasts in the career tra- jectory across different disciplines (Simonton, 1997). For instance, because the ideation and elaboration rates are faster for poetry than for novels, the age curve for poets peaks earlier and exhibits a more accelerated post-peak decline (Simonton, 1989). The com- binatorial model can even account for cross-sectional variation in career trajectories. This provision ensues from two individual- Correspondence concerning this article should be addressed to Dean Keith Simonton, Department of Psychology, One Shields Avenue, Univer- sity of California, Davis, CA 95616-8686. E-mail: dksimonton@ ucdavis.edu Psychology of Aesthetics, Creativity, and the Arts Copyright 2007 by the American Psychological Association 2007, Vol. 1, No. 3, 133–139 1931-3896/07/$12.00 DOI: 10.1037/1931-3896.1.3.133 133

Upload: kaybu

Post on 16-Nov-2014

245 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Simonton (2007) - Creative Life Cycles

Creative Life Cycles in Literature: Poets Versus Novelists orConceptualists Versus Experimentalists?

Dean Keith SimontonUniversity of California, Davis

The economist Galenson (2005) proposed a theory of creative life cycles that distinguishes betweenearly-peaking conceptual creators (finders) and late-peaking experimental creators (seekers). This con-trast is claimed to invalidate previous research findings that poets tend to peak earlier than novelists.However, a multiple regression analysis of his published data on 23 creative writers shows that thepoetry-novel genre contrast makes a contribution to the prediction of the career trajectory that isorthogonal to the conceptual-experimental contrast. The result is a fourfold typology of creative lifecycles: conceptual poets, conceptual novelists, experimental poets, and experimental novelists who dotheir best work at ages 28, 34, 38, and 44, respectively. The article closes with a discussion of additionalempirical and theoretical issues.

Keywords: creativity, age, life cycles, literary genres, conceptual versus experimental creators

Beyond doubt, the oldest topic in the empirical study of cre-ativity is the relation between age and creative output (Simonton,1999a). The first scientific analysis regarding this question wasconducted by Quetelet’s (1835/1968) concerning longitudinalchanges in the dramatic production of eminent British and Frenchplaywrights. About a century later the creative life cycle was takenup by Lehman in a series of articles that culminated in the 1953classic Age and Achievement. More recent research on this subjecthas been conducted by numerous other social and behavioralscientists (e.g., Cole, 1979; Dennis, 1966; Over, 1989; Stern, 1978;Simonton, 1991a; Zickar & Slaughter, 1999). In addition, it hasbeen the subject of comprehensive literature reviews (e.g.,Lindauer, 2003; Simonton, 1988a, 1996). Not surprisingly, giventhe sheer mass of research, many key results have been replicatedso many times as to constitute among the most secure articles ofknowledge in the psychological sciences.

Two findings are especially robust (Lehman, 1953; Dennis,1966; Simonton, 1988a, 1997).

First, the output of creative products tends to change over time,rising relatively quickly to a career maximum and then decliningsomewhat gradually thereafter. Typically, the peak occurs some-time in the late 30s or early 40s, and the productivity toward theend of the career is about half that at the career maximum. Thislongitudinal trend is approximated by an “inverted backward-J”function (Simonton, 1988a). This function is specified by asecond-order polynomial in age where the linear term is positiveand the quadratic term is negative.

Second, the specific shape of this single-peak function variesaccording to the domain of creative achievement. For instance, theoptimal age for writing poetry tends to be somewhat younger thanthat for writing novels (Dennis, 1966; Lehman, 1953; Simonton,

1975). Indeed, Simonton (1975) found that the age gap betweenpoets and novelists was invariant across history (from ancient tomodern times) and diverse literary traditions (Japanese, Chinese,Islamic, European, etc.). This effect is large enough to help explainwhy eminent poets tend to have a shorter life span than donovelists (Kaufman, 2003; Simonton, 1974; cf. Kaun, 1991). Be-cause poets produce their best work at a younger age, they can dieyounger without that event imposing a severe cost on their repu-tation (cf. McCann, 2001). A similar effect holds for mathemati-cians who display both accelerated career trajectories and shorterlife expectancies (Simonton, 1991a).

The age-output results have been the subject of several theoret-ical interpretations (e.g., Beard, 1874; Diamond, 1984; Simonton,1997; Stephan & Levin, 1992). Some of these explanations aresociological or economic, whereas others are clearly psychologicalin nature. In the latter category is a two-step combinatorial modelof the creative process that has been developed in a series ofarticles (Simonton, 1984, 1989, 1991a, 1997) and books (Simon-ton, 1988b, 2004a). In simple terms, this mathematical modelassumes that creators launch their careers with a certain amount ofcreative potential defined as a set of ideas or mental elementsavailable for free combination. This potential then undergoes thetwo-step process of ideation, by which useful combinations aregenerated, and elaboration, by which those combinations are con-verted into overt products. This two-step cognitive procedure thenyields a double-exponential age function that accurately predictsthe career trajectories indicated in previously published data sets(Simonton, 1984, 1989, 1997). Furthermore, because the ideationand elaboration rates are assumed to be domain-specific, the modelprovides a theoretical explanation for contrasts in the career tra-jectory across different disciplines (Simonton, 1997). For instance,because the ideation and elaboration rates are faster for poetry thanfor novels, the age curve for poets peaks earlier and exhibits amore accelerated post-peak decline (Simonton, 1989). The com-binatorial model can even account for cross-sectional variation incareer trajectories. This provision ensues from two individual-

Correspondence concerning this article should be addressed to DeanKeith Simonton, Department of Psychology, One Shields Avenue, Univer-sity of California, Davis, CA 95616-8686. E-mail: [email protected]

Psychology of Aesthetics, Creativity, and the Arts Copyright 2007 by the American Psychological Association2007, Vol. 1, No. 3, 133–139 1931-3896/07/$12.00 DOI: 10.1037/1931-3896.1.3.133

133

Page 2: Simonton (2007) - Creative Life Cycles

difference parameters, namely, the person’s initial creative poten-tial and the age at career onset (i.e., when the combinatorialprocess begins). As a consequence, the model successfully predictsthe key features of individual differences and longitudinal changesin creative output (Simonton, 1997). Many of these predictions areso distinctive to the model that their empirical confirmation rulesout several alternative explanations (e.g., any interpretation thatdefines the life cycle in terms of chronological rather than careerage).

Nonetheless, both this specific theoretical model and the generalempirical findings face a direct challenge from an unlikely quar-ter—from economics rather than psychology. This challenge ispresented in the 2005 Old Masters and Young Geniuses: The TwoLife Cycles of Artistic Creativity by David Galenson, an economistat the University of Chicago.1 The author put forward an alterna-tive theory of output trajectories in the arts and other creativedisciplines. His theory is predicated on the distinction between twoapproaches to creativity, the conceptual and the experimental. Inthe former category are the young geniuses or finders who con-ceive their best work as sudden intellectual breakthroughs thatappear very early in their careers. In the latter category are the oldmasters or seekers who work painstakingly via trial and error, andthus whose best work does not emerge until late in life when theycan finally savor the fruits of their labors. According to Galenson,Picasso represents the typical conceptual creator, whereas Cezanneexemplifies the experimental creator. Although the theory wasoriginally formulated with respect to painting (Galenson, 2001,Galenson (2005) extended it to encompass all major forms ofartistic creativity, including sculpture, poetry, novels, and film.Indeed, he has even extended the distinction to cover scientificcreativity (Weisberg & Galenson, 2005). Whatever the specificdomain of creative behavior, the conceptual finders have identifi-ably different work habits than the experimental seekers.

Galenson (2005) affirms that his theory is not just complemen-tary to earlier research results and theories. On the contrary, heclaims that his conception of the creative process supersedes theearlier psychological findings and interpretations (see especiallypp. 171–177). Rather than the age curves being a function of thedomain of creative activity, the trajectories are the upshot of aparticular approach to the creative process. Conceptualists peakearly and experimentalists peak late. Any apparent tendency forpoets to peak earlier than novelists may merely reflect differentialproportions of finders and seekers across the two literary genres(Galenson, 2006). If so, then controlling for the conceptualistversus experimentalist distinction would make the poet-novelistdifference disappear.

However, his argument is not immune from criticism. First,Galenson’s judgments of the psychological literature exhibit somemisunderstandings of the research. For instance, he said that “thegeneralizations of the psychologists quoted here [regarding thepoetry-novel age contrast] may stem from an uncritical acceptanceof the findings of Lehman, buttressed by the dramatic examples ofByron, Keats, Shelley, and the other famous young geniuses whodied prematurely” (p. 176). This statement overlooks the fact thatthe age gap has been established for hundreds of literary creatorsrepresenting all of the world’s major literatures, and in analysesthat implemented statistical controls for such contaminants as lifespan (Simonton, 1975). Moreover, psychologists do not uncriti-cally claim that all creators have identical career trajectories but

only that the most typical or average curve is roughly described bythe inverted backward-J curve (see, e.g., Zickar & Slaughter,1999). A minority of creators can depart appreciably from thisoverall trend without threatening the generalization. Accordingly,there is no logical reason for adopting an either-or stance. Becausepsychologists recognize individual differences in longitudinaltrends, Galenson’s theory can be used to explain those instances inwhich some creators peak earlier or later than the norm. The earlybloomers may be conceptualists and the late-bloomers experimen-talists. Yet given that the overall curve remains single-peakedrather than bimodal, these departures would have to represent theexceptions rather than the rule (Cole, 1979; Dennis, 1966; Leh-man, 1953; Quetelet, 1835/1968). The repeatedly replicated uni-modal distribution could then indicate that most creators adopt acombination of conceptual and experimental styles.

A more fundamental problem concerns the methodology under-lying Galenson’s (2005) substantive conclusions. Few of his stud-ies on this specific issue are published in refereed journals. Rather,the vast majority appear as “working papers” deposited at theNational Bureau for Economic Research (http://papers.nber.org/papers/). As a result, his research sometimes lacks the method-ological rigor expected in peer-reviewed publications. For in-stance, Winner (2004), commenting on an earlier presentation ofthe theory (Galenson, 2001), observed that his operational defini-tion of the core seeker-finder distinction appears too vague andeven contradictory. Thus, although Galenson classified Picasso asa conceptualist painter, the manner in which that artist conceivedGuernica is far more indicative of an experimental painter (seealso Simonton, in press). At times Galenson appears to bifurcate agroup of creators into early- and late-bloomers and then try toidentify possible differences between the two groups—a procedurethat confounds the operational definition with the empirical test ofthe theory.

Furthermore, in contrast with the psychological research that hecriticizes, Galenson (2005) does not usually apply statistical anal-yses that would enable him to (a) control for potential artifacts and(b) obtain unbiased estimates of effect sizes. Most often he simplyprovides tables and graphs depicting group differences in whichthe conclusions must be based on visual inspection. In addition, thesamples on which he founds his inferences are frequently smallerthan those seen in most psychological research (i.e., dozens ratherthan hundreds). Finally, the sampling procedures are often vaguelydefined for any given investigation and inconsistent across sepa-rate investigations. Consequently, samples recurrently appear notjust small but also arbitrary. All in all, Gardner (2006) may havebeen justified when he held that “Galenson. . .is not. . .a genuineempiricist, but rather a theorist who is posing as an empiricist” (p.278).

But for our current purposes the most substantial objection maybe that Galenson’s (2005) research is fragmented rather thanintegrated. Rather than introduce a single method—sampling pro-

1 I should probably point out that I served as a referee when the bookmanuscript was submitted to Princeton University Press. Besides recom-mending publication, I consented to provide a quote for the book’s dustjacket. The blurb read “A very well written and intellectually stimulatingpiece of scholarship that deserves to be widely read and debated.” It is inthat spirit that I contribute the current piece.

134 SIMONTON

Page 3: Simonton (2007) - Creative Life Cycles

cedures, variable definitions, and statistical analyses—to inclusivesamples of creators, he favors methodologically variable inquiriesinto more narrow samples. For example, despite his strong objec-tion to the generalizations of the psychological research on thepoetry-novel difference in peak ages, he has never conducted aninvestigation looking at both poets and novelists together. Instead,he has put out one working paper on modern American poets(Galenson, 2003) and another on modern novelists (Galenson,2004). In neither paper does he make direct numerical comparisonswith the findings of the other paper. That means that he ignoreswhether the effect of literary genre operates as a determinant ofcreative life cycle independent of the impact of the contrast be-tween finders and seekers. Yet, as already noted, these two factorsneed not be mutually exclusive. A more inclusive alternative is thatfour distinct career trajectories operate, namely, finder poets,seeker poets, finder novelists, and seeker novelists.

To detect this second outcome requires that the two samples beincorporated into a single quantitative investigation. Fortunately,because Galenson (2003, 2004) published the raw data, this inte-grative analysis can be performed here. Are there four careertrajectories, or just two?

Method

Sample

The 23 writers in this study were divided into 11 poets fromGalenson (2003) and 12 novelists from Galenson (2004). Therewere three female poets (Elizabeth Bishop, Marianne Moore, andSylvia Plath) and one female novelist (Virginia Woolf). Althoughthe poets are all from the United States, the novelists came fromboth the United States and Great Britain (counting Joseph Conradas British rather than Polish). The birth years are even moreheterogeneous. Despite the fact that both sets of writers wereidentified as “modern,” the poets were born between 1874 (RobertFrost) and 1932 (Sylvia Plath), and the novelists between 1812(Charles Dickens) and 1899 (Ernest Hemmingway), making thepoets more recent than the novelists. The difference between themean birth years was about 35 years. However, this differenceactually biases the data against finding the expected poet-novelistage gap. Because the lifespans of eminent creators have beengradually increasing over historical time (e.g., Simonton, 1977)and because longer lifespans are positively correlated with latercareer peaks (Lindauer, 1993; Simonton, 1975, 1991a, 1991b), thenovelists should have their optima shifted toward earlier ages. Inthe present sample, the average life span of the poets was 72.1year, that of the novelists 63.9 years, a gap of 8.2 years that shouldreduce the difference in career peaks.

Measures

The dependent variable is the age at which the writer wrote hisor her single best work. For the poets the best work was defined asthe most frequently anthologized poem (according to Tables 5 and7 of Galenson, 2003). For the novelists the best work was deter-mined by the amount of space devoted to each work in 10 criticalmonographs devoted to each author (according to Tables 4 and 5of Galenson, 2004). Although these two operational definitions arenot equivalent, prior research shows that alternative archival indi-

cators of esthetic merit exhibit high intercorrelations, a finding thatholds for drama, poetry, opera, and film (Simonton, 1986, 1990,1998, 2004b). So let us call this criterion Age best work.

There are two substantive predictors, one capturing the claims ofpast psychological research and the other Galenson’s contraryposition. The first is a dummy variable recording the domain ofliterary creativity. In particular, the variable Novelist equaled 1 ifthe writer was in the Galenson (2004) sample, but equaled 0 if inthe Galenson (2003) sample. Although poets can be novelists,these two samples did not overlap, so that there was no ambiguityin the definition of this variable. For example, although ThomasHardy was a notable English poet, he was not an American, and sohe could not have been included in Galenson (2003) study. Ac-cordingly, Hardy could be categorized exclusively as a novelist.Most likely, this is how he would have been identified anyway ifthe assignment was based on Hardy’s single greatest work, namelythe novel Tess of the d’Urbervilles.

The second substantive predictor was the dummy variable Ex-perimentalist that equaled 1 if the writer was labeled such byGalenson (2003, 2004) and equaled 0 if identified as a conceptu-alist. Almost 57% of the writers across both literary genres wereclassified as experimental creators, the rest being conceptual cre-ators. It should be noted that the manner in which Galensonderived these classifications was not identical for the novelists andpoets. In the former case some resemblance of an operationaldefinition played a greater role in the categorization, whereas inthe latter case he seems to have worked backward from the ages atwhich the poets produced their best work.

The last independent variable is a statistical control variable,namely, the writer’s life span. Because one poet was still living(Richard Wilbur), he was assigned an artificial life span by havinghim die in 2006, at the time of this investigation (viz., age 85). Thisappears reasonable given that many years have lapsed since thepoet has made a major contribution to the genre (i.e., his last majorbook of poems appeared in 1988). With this inserted value thelifespans ranged from 31 (Sylvia Plath) to 89 (Robert Frost) witha mean of 68.4. To render the regression results more interpretable,this variable was put in mean-deviation form.

Results

The analysis consisted of two consecutive multiple regressions.In Model 1 the dependent variable (age best work) was regressedon both substantive variables (Novelist and Experimentalist dum-mies) and in Model 2 the control variable (life span) was added tothe equation. Table 1 shows the outcome for both models.

Model 1

In the first regression equation the intercept indicates the ex-pected age at best work for those creative writers who were neithernovelists nor experimentalists—namely, the conceptualist poets.This group writes their best poem around age 28. The unstandard-ized partial regression coefficient for the first dummy variable(Novelist) then indicates whether being a conceptual novelistchanges this expectation. According to the results, these novelistsare about 5 years older when they produce their best work. Moreprecisely, they tend to be 33.53 (� 28.27 � 5.26) when their bestnovel appears. The unstandarized partial regression coefficient for

135CREATIVE LIFE CYCLES

Page 4: Simonton (2007) - Creative Life Cycles

the second dummy variable (Experimentalist) similarly indicateswhether being an experimental poet alters the expected age relativeto being a conceptual poet. In this case the increment is almost 11years. More specifically, experimental poets produce their bestwork at age 38.94 (� 28.27 � 10.67), or about 5 years older thanconceptual novelists. The final contingency is when a writer is anexperimental novelist, in which case the predicted age at best workis obtained by adding all three unstandardized coefficients. Inparticular, experimental novelists are expected to do their bestwork at age 44.20 (� 28.27 � 5.26 � 10.67). Taken together thesetwo dummy variables account for 60% of the variance in the agethat creators do their best work. This fit to the model is excep-tionally good (cf. Simonton, 1975).

It is thus clear that both factors make substantively and statis-tically significant contributions to the prediction of the career peak.The effect size for genre is comparable to that found in earlierinvestigations. For instance, Simonton’s (1975) study of creativewriters obtained an age difference of about 4 years. At the sametime, the effect of the seeker-finder distinction is about double thatof genre (i.e., 2.03 � 10.67/5.26). This greater impact is alsoindicated by comparing the two standardized partial regressioncoefficients; the Experimentalist coefficient is exactly double thatfor Novelist (i.e., 2 � .68/.34). Nevertheless, it must be empha-sized that the two factors make practically orthogonal contribu-tions to predicting the creative life cycle. That independence isshown by the fact that the correlation between the two dummyvariables is practically zero (i.e., r � .04, p � .8627). Morespecifically, while there is a very slight tendency for poets to beconceptualists and novelists to be experimentalists, this associationis neither substantively nor statistically significant. For the mostpart, conceptual and experimental writers are evenly distributedamong the two literary genre, at least according to Galenson’s(2003, 2004) own data.

Model 2

Nonetheless, the results for Model 1 may be contaminated withthe influence of life span. This possibility emerges from the factthat life span correlates with the dependent variable and bothindependent variables. The specific correlations are: Age bestwork .24, Novelist .32, and Experimentalist �.32. Admittedly,because we are dealing with a very small sample, none of thesecorrelation coefficients are statistically significant (viz., ps of.2795, .1358, and .1406, respectively). Even so, because they

indicate that the amount of shared variance is around 10%, thecorrelations are large enough to have substantive consequences,including suppression effects. This outcome is evident in theModel 2 statistics presented in Table 1.

Despite the fact that life span is not a significant predictor of ageat best work, life span does affect appreciably the magnitude of thepoet-novelist and finder-seeker contrasts. Specifically, the effect ofNovelist is increased by almost a year to a bit over 6 years whilethe effect of Experimentalist is decreased by almost exactly thesame amount to a bit less than 10 years. As a consequence, the twoeffect sizes are more equal, albeit the conceptual-experimentalcontrast still surpasses that for poet-novelist. In any case, theserevised coefficients can be combined with the slightly changedintercept to obtain new predicted scores for the four categories ofliterary creativity: conceptual poets 28.32, conceptual novelists34.38 (� 28.32 � 6.06), experimental poets 38.17 (� 28.32 �9.85), and experimental novelists 44.23 (� 28.32 � 6.06 � 9.85).In more approximate terms, we obtain four distinct career peaks inthe late 20s, the middle 30s, the late 30s, and the middle 40s.Almost 16 years separates the conceptual poets from the experi-mental novelists.

It should now be apparent that there are not just two kinds ofcreative life cycles, as Galenson (2005) argued, but rather at leastfour. Furthermore, the near-zero correlation between the two dum-mies indicates that these four can be considered orthogonal types.The nature of this typology is indicated in Table 2 where the 23writers are grouped according to placement along the two dimen-sions. Within each category the creators are ordered according tothe age at which they produced their best work—their individualcareer peaks. Moreover, at the top of each quadrant are the pre-dicted career peaks for each of the four types (using Model 2 fromTable 1). It is manifest that the observed scores tend to be scatteredaround the corresponding predicted scores. About half of thewriters have earlier peaks than expected and about half have laterpeaks than expected, with most having peaks fairly close to pre-diction. In addition, it is evident that the conceptualists tend topeak earlier than the experimentalists, with only modest overlapbetween the distributions. Yet it is also true that the poets tend topeak earlier than do the novelists. This is very apparent if youcompare each poet with each novelist lined up in the same row: Inalmost every case the poet has an earlier peak than the correspond-ing novelist or novelists. This consistent ordinal difference showsthat the genre distinction is not trivial.

Table 1Multiple Regression Analysis: Predictors of Age Best Work for 23 Poets and Novelists

Style

Model 1 Model 2

b � p b � p

Intercept (Poet/Conceptualist) 28.27 .00 .0000 28.32 .00 .0000Novelist 5.26 .34 .0274 6.06 .39 .0187Experimentalist 10.67 .68 .0001 9.85 .63 .0006Lifespan 0.08 .16 .3379R2 .60 .0001 .62 .0003

Note. The intercept is the predicted age when both Novelist � 0 and Experimentalist � 0. The statistics provided are the unstandardized partial regressioncoefficient (b), the standardized partial regression coefficient (�), the probability level of the t test ( p), and the multiple correlation squared (R2), whichgives the total proportion of variance explained.

136 SIMONTON

Page 5: Simonton (2007) - Creative Life Cycles

Discussion

According to Galenson’s (2003, 2004) own data, the distinctionbetween poetry and novels remains a critical factor in understand-ing creative life cycles in literature. This is not to say that thedomain of creativity is just as important as the style of creativity.After all, the seeker-finder distinction has a noticeably largerimpact on age at best work. Still, the literary genre adds a signif-icant increment to our predictive power. In the case of Model 1 (inTable 1), for example, the Experimental dummy variable alonewould account for 48% of the variance. Adding the Novelistdummy variable to the equation increases that proportion to 60%.If science were a game of finding the single best predictor, thefinder-seeker variable would be at present the predictor of choice.But science is not engaged in contests of this kind. The primarygoal is to obtain a complete comprehension of a phenomenonusing as many factors as necessary. As Table 2 shows, includingboth factors provides a richer understanding of creative life cyclesin literature.

To appreciate better the implications of these results I wouldlike to address two sets of issues, one empirical and the othertheoretical.

Empirical Issues

As an empirical investigation this study leaves much to bedesired. First and foremost, the sample size was much smaller thanwould be normally recommended. There were only 23 cases aboutevenly divided between poets and novelists. This contrasts greatlywith the number of cases in previous psychological research on theage-output curve. For instance, Simonton (1975) examined 420literary creators, and Simonton (1991a) has tested his model ofinterdisciplinary differences in career trajectories on over 2,000scientists. Even Dennis (1966), who gathered somewhat smallersamples than usual, included 46 poets and 32 novelists, over threetimes the size of Galenson’s (2003, 2004) combined samples.

There are two main reasons why this issue should be addressedwith larger Ns than Galenson tends to use. First, small samples

make it more difficult to reject the null hypothesis (i.e., they havelower statistical power). As a result, it is possible that some moresubtle determinants of creative life cycles will be overlookedbecause of Type II errors. Second, a fuller appreciation of thephenomenon requires more complex models than those imple-mented in the current investigation. For example, Simonton (1975)included 23 independent variables in his study of the career peaksof 420 writers—as many variables as the number of cases in thecurrent analysis! Because N should amply exceed the number ofpredictors, a small sample size restricts the sophistication of themodel that can be tested.

The composition of the sample is also problematic. In thepresent secondary analysis the novelists formed a more heteroge-neous group than the poets both with respect to historical periodand geographical origins. Hence, the two groups were not equiv-alent. Either the novelists should have been confined to modernAmerican novelists or the American poets should have been ex-panded to include all modern poets. Of course, the latter course ofaction is probably best. By increasing the diversity of the samplewe can determine whether any observed differences are invariantacross time and place (Lehman, 1953; Simonton, 1975, 1991a,1991b).

In addition, it is imperative to adopt a more uniform samplingstrategy. As pointed out earlier, sometimes Galenson (2005) con-founds the selection strategy with his theoretical concepts. Thisproblem is apparent in his sample of poets. An inspection of Table2 reveals that not a single poet produced their best work betweenthe ages of 35 and 39 inclusively. This hiatus is perplexing becauseit represents the half decade in which the best work has a highprobability of appearing. For instance, according to one transhis-torical and cross-cultural study, the average age that poets pro-duced their single most important composition was 38.8(Simonton, 1975). And no investigation prior to Galenson’s (2003)has identified a bimodal distribution with a zero-output troughsituated in this age interval (see, e.g., Lehman, 1953). Therefore,insofar as the work of the “average poet” has been omitted fromthe analysis, it is likely that the effect size for the seeker-finder

Table 2Typology of Creative Life Cycles in Literature: Predicted and Observed Career Peaks

Style Poets Novelists

Conceptualists (finders) Predicted: 28 Predicted: 34Eliot (1888 � 1965): 23 Fitzgerald (1896 � 1940): 29Cummings (1894 � 1962): 26 Hemingway (1899 � 1961): 30Plath (1932 � 1963): 30 Melville (1819 � 1891): 32Pound (1885 � 1972): 30 Lawrence (1885 � 1930): 35Wilbur (1921�): 34 Joyce (1882 � 1941): 40Williams (1883 � 1963): 40

Experimentalists (seekers) Predicted: 38 Predicted: 44Bishop (1911 � 1979): 29 James (1843 � 1916): 38Moore (1887 � 1972): 32 Faulkner (1897 � 1962): 39Lowell (1917 � 1977): 41 Dickens (1812 � 1870):Stevens (1879 � 1955): 42 Woolf (1882 � 1941): 45Frost (1874 � 1963): 48 Conrad (1857 � 1924): 47

Twain (1835 � 1910): 50Hardy (1840 � 1928): 51

Note. The predicted career peak (age at single best work) was generated from Model 2 in Table 1 (rounded off to the nearest integer). Within each typeare listed the writers in Galenson (2003, 2004) according to their observed career peaks (given after the colon).

137CREATIVE LIFE CYCLES

Page 6: Simonton (2007) - Creative Life Cycles

distinction is biased upward. The true effect may be closer to thatof genre and could even be smaller.2

A final methodological issue has already been mentioned in theintroduction section: The need for a precise operational definitionof what counts as conceptual versus experimental creativity(Gardner, 2006; Winner, 2004). Yet it turns out that this is not justa methodological issue but a theoretical problem besides.

Theoretical Issues

In Simonton’s (1984, 1997) two-step model the ideation andelaboration rates that determine the career course are domainspecific. That is, the two rates are based on the particular nature ofthe domain’s concepts, techniques, standards, and modes of ex-pression or communication (Simonton, 2004a). The consequenceof these features is then the distinctive life cycle that characterizeseach domain. This does not mean that creators working within thesame domain will not vary in their trajectories. On the contrary, thetwo-step model actually includes two variables to deal with indi-vidual differences, and the model also allows for the operation ofdiverse “random shocks” that deflect a particular creator’s outputfrom the expected trajectory (Simonton, 1997, 2004a). Even so,domain-specific conditions play a major role in determining thecreator’s overall career path.

Galenson (2005), in contrast, believes that the creative life cycleresides in the individual creator. Some happen to be conceptualfinders, others experimental seekers—and this distinction so dom-inates the career course that domain differences become irrelevant.It is for this reason that Galenson can dismiss the psychologicalinvestigations that document domain-specific career peaks. Ironi-cally, in a sense Galenson’s theory is more psychological thanSimonton’s because the former places the causal locus within theindividual creator rather than outside in the domain. Yet when wemore closely scrutinize Galenson’s theoretical position, we en-counter two problems.

First, if seekers and finders are manifestations of underlyingpsychological factors, it is difficult if not impossible to considerthis distinction as categorical or even bimodal. Virtually everypsychological construct ever measured is unimodal. Most often thetrait is distributed in a close approximation to the normal orbell-shaped curve, but sometimes the characteristic will display ahighly skewed but still unimodal distribution (Simonton, 1999b).Although psychologists may loosely speak of, say, extrovertsversus introverts, it is commonly understood that the true extrovertor introvert is found on the extreme tail of the distribution. Mostindividuals fall in the middle, sometimes being extroverted andother times introverted. Given this fundamental reality of humanindividual-difference variables, pure conceptualists and pure ex-perimentalists must be extremely rare should they have a psycho-logical foundation. The overwhelming majority of creators wouldhave to represent some mix of the two strategies. Notice that thisproblem is not found in any model that attributes the life cycles todomain characteristics. The properties of domains are discreterather than continuous, multimodal rather than unimodal. Certainlythe attributes of poetry are distinct from the attributes of novels.Creative products in the two genres aspire to accomplish verydifferent literary ends with rather distinctive means. Indeed, theextreme lack of commonalities between the domains may account

for the comparative rarity of writers who have managed to producefirst-rate works in both genres.

Second, when Galenson (2005) endeavors to root the distinctionin the individual creator, his description of the conceptual-experimental contrast contains numerous references to domain-specific traits. To illustrate, he specifies the difference between thetwo kinds of poetry as follows:

whereas conceptual poetry often involves introspection, experimentalpoetry typically involves observation. Conceptual poetry often growsout of a study of earlier poetry, whereas experimental poetry moreoften comes from study of the external world; conceptual poets mayfind their raw material in libraries, but experimental poets are morelikely to find it by traveling or working at other professions. Concep-tual poetry is often concerned primarily with technique, whereasexperimental poetry tends to emphasize subject matter. Conceptualpoetry is more often abstract, and aimed at universality, while exper-imental poetry is generally concrete, and concerned with specifics.The language of conceptual poetry is more likely to be formal orartificial, while that of experimental poetry may be informal andvernacular. Conceptual poetry is more often based on imagination,experimental poetry on the author’s perception of reality. (Galenson,2003, pp. 8–9)

Many of these identified differences are among those thatSimonton (1984, 1989, 1991a, 1997, 2004a) used to character-ize the domain-specific contrasts that influence the ideation andelaboration rates. In particular, domains with fast ideation ratesdeal with a more limited and well-defined inventory of abstractconcepts that are highly constrained by a system of rules,whereas those with slower ideation rates deal with more unlim-ited and ill-defined inventory of concrete concepts that are moreheavily dependent upon accumulated experience. These distinc-tions are important because they moderate the speed of thecombinatorial process. In addition, subtle changes in domainattributes can exert large shifts in the expected creative lifecycles (Simonton, 1989). These effects are even apparent withina specific domain like poetry. Despite the common tendency totreat all poetry as representing the same domain, this form ofcreativity should be differentiated into several subdomains,each with their separate expected career trajectories (Lehman,1953; Simonton, 1975). Examples include such categories aslyric, pastoral, narrative, dramatic, epic, satiric, humorous, po-litical, religious, and didactic—not even counting such specificforms as sonnets, odes, songs, and even limericks. Some ofthese have early predicted peaks and others late predictedpeaks, all depending on the distinctive mix of attributes for thatparticular poetic expression.

Hence, aside from all of the methodological issues that shouldbe addressed, Galenson should develop a more complete theoret-ical basis for his distinction. This development would require (a)an indication of the psychological variables that determine whethersomeone becomes a finder or a seeker and (b) a specification of

2 Because the same gap is not found for the novelists, this assertionimplies that the finder-seeker effect might be stronger for the poets. Thispossibility was examined by introducing into both models a multiplicativeterm for the Experimentalist � Novelist interaction effect. In neither casewas the interaction term statistically or substantively significant. However,statistical tests for interaction effects have low power in small samples. Sothe result may be considered inconclusive.

138 SIMONTON

Page 7: Simonton (2007) - Creative Life Cycles

why any finder-seeker differences cannot be attributed to domain-specific features of the creative process.

References

Beard, G. M. (1874). Legal responsibilty in old age. New York: Russell.Cole, S. (1979). Age and scientific performance. American Journal of

Sociology, 84, 958–977.Dennis, W. (1966). Creative productivity between the ages of 20 and 80

years. Journal of Gerontology, 21, 1–8.Diamond, A. M., Jr. (1984). An economic model of the life-cycle research

productivity of scientists. Scientometrics, 6, 189–196.Galenson, D. W. (2001). Painting outside the lines: Patterns of creativity

in modern art. Cambridge, MA: Harvard University Press.Galenson, D. W. (2003). Literary life cycles: The careers of modern

American poets. Working Paper 9856, National Bureau of EconomicResearch. Retrieved August 29, 2006, from http://papers.nber.org/papers/

Galenson, D. W. (2004). A portrait of the artist as a young or old innovator:Measuring the careers of modern novelists. Working Paper 10213,National Bureau of Economic Research. Retrieved August 29, 2006,from http://papers.nber.org/papers/

Galenson, D. W. (2005). Old masters and young geniuses: The two lifecycles of artistic creativity. Princeton, NJ: Princeton University Press.

Galenson, D. W. (2006, March 10). A portrait of the artist as a very youngor very old innovator: Creativity at the extremes of the life cycle. Paperpresented at the New Ideas About New Ideas Conference, NationalBureau of Economic Research, Cambridge, MA.

Gardner, H. (2006). Replies to my critics. In J. A. Schaler (Ed.), HowardGardner under fire: A rebel psychologist faces his critics (pp. 245–312).Chicago: Open Court.

Kaufman, J. C. (2003). The cost of the muse; poets die young. DeathStudies, 27, 813–821.

Kaun, D. E. (1991). Writers die young: The impact of work and leisure onlongevity. Journal of Economic Psychology, 12, 381–399.

Lehman, H. C. (1953). Age and achievement. Princeton, NJ: PrincetonUniversity Press.

Lindauer, M. S. (1993). The span of creativity among long-lived historicalartists. Creativity Research Journal, 6, 231–239.

Lindauer, M. S. (2003). Aging, creativity, and art: A positive perspectiveon late-life development. New York: Kluwer Academic/Plenum PressPublishers.

McCann, S. J. H. (2001). The precocity-longevity hypothesis: Earlier peaksin career achievement predict shorter lives. Personality and SocialPsychology Bulletin, 27, 1429–1439.

Over, R. (1989). Age and scholarly impact. Psychology and Aging, 4,222–225.

Quetelet, A. (1968). A treatise on man and the development of his faculties.New York: Franklin. (Reprint of 1842 Edinburgh translation of 1835French original).

Simonton, D. K. (1974). The social psychology of creativity: An archivaldata analysis. Unpublished doctoral dissertation, Harvard University.

Simonton, D. K. (1975). Age and literary creativity: A cross-cultural andtranshistorical survey. Journal of Cross-Cultural Psychology, 6, 259–277.

Simonton, D. K. (1977). Eminence, creativity, and geographic marginality:A recursive structural equation model. Journal of Personality and SocialPsychology, 35, 805–816.

Simonton, D. K. (1984). Creative productivity and age: A mathematical

model based on a two-step cognitive process. Developmental Review, 4,77–111.

Simonton, D. K. (1986). Popularity, content, and context in 37 Shakespeareplays. Poetics, 15, 493–510.

Simonton, D. K. (1988a). Age and outstanding achievement: What do weknow after a century of research? Psychological Bulletin, 104, 251–267.

Simonton, D. K. (1988b). Scientific genius: A psychology of science.Cambridge: Cambridge University Press.

Simonton, D. K. (1989). Age and creative productivity: Nonlinear estima-tion of an information-processing model. International Journal of Agingand Human Development, 29, 23–37.

Simonton, D. K. (1990). Lexical choices and aesthetic success: A computercontent analysis of 154 Shakespeare sonnets. Computers and the Hu-manities, 24, 251–264.

Simonton, D. K. (1991a). Career landmarks in science: Individual differ-ences and interdisciplinary contrasts. Developmental Psychology, 27,119–130.

Simonton, D. K. (1991b). Emergence and realization of genius: The livesand works of 120 classical composers. Journal of Personality and SocialPsychology, 61, 829–840.

Simonton, D. K. (1996). Creativity. In J. E. Birren (Ed.), Encyclopedia ofgerontology (pp. 341–351). San Diego, CA: Academic Press.

Simonton, D. K. (1997). Creative productivity: A predictive and explana-tory model of career trajectories and landmarks. Psychological Review,104, 66–89.

Simonton, D. K. (1998). Fickle fashion versus immortal fame: Transhis-torical assessments of creative products in the opera house. Journal ofPersonality and Social Psychology, 75, 198–210.

Simonton, D. K. (1999a). Significant samples: The psychological study ofeminent individuals. Psychological Methods, 4, 425–451.

Simonton, D. K. (1999b). Talent and its development: An emergenic andepigenetic model. Psychological Review, 106, 435–457.

Simonton, D. K. (2004a). Creativity in science: Chance, logic, genius, andzeitgeist. Cambridge, England: Cambridge University Press.

Simonton, D. K. (2004b). Group artistic creativity: Creative clusters andcinematic success in 1,327 feature films. Journal of Applied SocialPsychology, 34, 1494–1520.

Simonton, D. K. (in press). The creative imagination in Picasso’s Guernicasketches: Monotonic improvements or nonmonotonic variants? Creativ-ity Research Journal.

Stephan, P. E., & Levin, S. G. (1992). Striking the mother lode in science:The importance of age, place, and time. New York: Oxford UniversityPress.

Stern, N. (1978). Age and achievement in mathematics: A case-study in thesociology of science. Social Studies of Science, 8, 127–140.

Weisberg, B. A., & Galenson, D. W. (2005). Creative careers: The lifecycles of Nobel laureates in economics. Working Paper 11799, NationalBureau of Economic Research. Retrieved August 29, 2006, from http://papers.nber.org/papers/

Winner, E. (2004, July 2). Art history can trade insights with the sciences.Chronicle of Higher Education, B10–12.

Zickar, M. J., & Slaughter, J. E. (1999). Examining creative performanceover time using hierarchical linear modeling: An illustration using filmdirectors. Human Performance, 12, 211–230.

Received August 31, 2006Revision received November 13, 2006

Accepted November 21, 2006 �

139CREATIVE LIFE CYCLES