Ascent in Competitive Arenas: from Fenway Park to Mass Ave

Ascent in competitive arenas: From Fenway Park to Mass Ave

The Science of Success: Measurements and Predictions June 17th Harvard University http://www.barabasilab.com/success! Follow us on Facebook for updates: www.facebook.com/SuccessScience

The availability of massive data on individual performance has prompted scientists to start exploring patterns that govern the path to individual success. The topic is diverse — from exploring citations to influential scientific papers to the emergence of runaway videos on YouTube, from the popularity of hash tags on Twitter to the path to success for countries. As such the tools and perspectives vary, engaging social scientists, computer scientists, economists, physicists, mathematicians, and many other disciplines. The goal of this symposium is to bring together these diverse communities. We invite contributions and participants in two tracks:

Ignite Talk Submit a one-page abstract. Selected participants in this track are invited to present an Ignite talk about their research — a five minutes talk accompanied by 20 slides. Slides are automatically advanced with 15 seconds each — A prize committee willAlexander judge and select M. Petersenthe best talk for an award sponsored by IBM Research. Submission should be emailed to [email protected], with subject indicating 'Ignite Talk'. The IMT email shouldInstitute include for the Advanced information about Studies, the presenter, Lucca including Italy name, current position and affiliation.

Guest If you wish to attend the event without giving a presentation, apply for a guest seat. Tell us a bit about yourself by emailing us a short note that indicates your motivation for participating as well as your name, position, and affiliation at [email protected], with subject indicating 'Guest'.

Submissions will be evaluated in a rolling basis. Evaluation will continue until all seats are filled. Space is limited, so those interested should submit their abstracts as soon as possible.

Submission Deadline: May 17, 2013. Acceptance Notification: May 24, 2013. There is no registration fee, but proof for acceptance to either of the two tracks is needed for participation.

Venue The Institute for Quantitative Social Science, Harvard University, Cambridge, MA, USA

Organized by: Albert-László Barabási, David Lazer, Dashun Wang Center for Complex Network Research, Northeastern University, Boston, MA, USA

Confirmed Speakers as of April, 2013 Pierre Azoulay, Sloan School of Management, MIT James A. Evans, Professor of Sociology, University of Chicago Santo Fortunato, Associate Professor in Complex Systems, Aalto University, Finland Gautam Mukunda, Harvard Business School Alexander M. Petersen, IMT Institute for Advanced Studies Lucca, Italy Nicola Perra, Northeastern University Chaoming Song, CCNR, Northeastern University Camille Sweeney, author of "The Art of Doing" Christoph Riedl, IQSS, Harvard University Arnout van de Rijt, Department of Sociology, SUNY Stony Brook Brian Uzzi, the Kellogg School of Management, Northwestern University Duncan Watts, Principal Researcher at Microsoft Research 103 Mozart x(t) = compositions stellar career growth is )

t 102 αi =1.42 (

i a non-linear process X

101 age 8

100 101 cumulative output ~ reputation Babe Ruth 101 x(t) = OPS t αi =1.24 ) αi t X (t) x (τ) t

( i i i ≡ ∼ X 100 τ=1 α 1 constant output, no growth 100 101 i 105 ≈ ⇒ H. Eugene Stanley super-linear growth 104 αi > 1 x(t) = annual citations ⇒ ) 103 t αi =2.41 ( i 102 X

101

100 100 101 career age, t Coevolution of reputation and impact over the academic career

Alexander M. Petersen,1 Santo Fortunato,2 Raj K. Pan,2 Orion Penner,3 Massimo Riccaboni,3, 4 H. Eugene Stanley,5 and Fabio Pammolli1, 5, 6 1Laboratory for the Analysis of Complex Economic Systems, IMT (Institutions Markets Technologies) Institute for Advanced Studies Lucca, 55100 Lucca, Italy 2Department of Biomedical Engineering and Computational Science, Aalto University, 12200 Aalto, Finland 3Laboratory of Innovation Management and Economics, IMT Institute for Advanced Studies Lucca, 55100 Lucca, Italy 4Department of Managerial Economics, Strategy and Innovation, Katholieke Universiteit Leuven, 3000 Leuven, Belgium 5Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215, USA 6Crisis Lab, IMT Lucca Institute for Advanced Studies, Lucca 55100, Italy (Dated: July 26, 2012) The cumulative citations to a paper is a universal measure of impact, but the role that author reputation plays in the life-cycle of the citation rate remains poorly understood. As a result, models of citation dynamics and career trajectories overlook the collaboration and reputation spillovers constitute a cumulative advantage underlying the competitive aspects of science. To better understand the reputation effect in science, we analyze the longitudinal citation dynamics of 350 leading scientists from biology, physics, and mathematics. We uncover statistical regularities in the evolution of productivity and impact which we use as benchmarks for a theoretical model of career growth that we test and validate on real careers. Our model incorporates the life-cycle effect for individual papers, the cumulative advantage arising from scientific reputation, and the preferential attachment effect for citation dynamics. We find that the author reputation effect dominates in the initial phase of the citation life-cycle, but that the preferential attachment mechanism emerges as the main component behind the sustained citation rate of highly cited papers. Comparing betweenCoevolution the three disciplines, of reputation we show and that impact the impact over life-cycle the academic career differs between fields: the axiomatic discoveries in mathematics have a very long shelf-life, whereas the rapid pace in biology and physics results in a shorter half-lifeAlexander despite the M. intense Petersen, citation1 Santo rate Fortunato, in the field.2 Raj K. Pan,2 Orion Penner,3 Massimo Riccaboni,3, 4 H. Eugene Stanley,5 and Fabio Pammolli1, 5, 6 1 Todo: by constructionLaboratory for (L thecorresponds Analysis of Complex to the Economic career length Systems, of in- IMT (Institutions Markets Technologies)i Institute for Advanced Studies Lucca, 55100 Lucca, Italy 2 dividual i at the time of data extraction). Fig. 1 shows 2 Department of Biomedical Engineering and Computational Science, Aalto University, 12200 Aalto, Finland Perform statistical χ significance tests on the DGBD 3 • the characteristicLaboratory of production Innovation Management trajectory and obtained Economics, by averag- profiles for datasets [D] and [E] and put in SI. ingIMT together Institute the forA Advancedindividual Studies trajectories Lucca, 55100N˜i(t Lucca,) belonging Italy to 4Department of Managerial Economics,A Strategy and Innovation, each dataset, N˜(t) Ni(t)/A ni , where we define Calculate πi, ρi, and τi for each of 350 scientists, put in Katholieke Universiteit Leuven,i=1 3000 Leuven, Belgium 5 ≡ • N (t)Center= N˜ for(t Polymer) / N˜(1) Studies. and Department of Physics, tables, and look for relations to other factors ThisBoston regularity University, reflects Boston, the Massachusetts abundance of 02215, of careers USA with 6Crisis Lab, IMT Lucca Institute for Advanced Studies, Lucca 55100, Italy αi > 1 corresponding(Dated: to accelerated July 26, 2012) career growth. This acceleration is consistent with increasing returns arising from I. INTRODUCTION The cumulativeknowledge citations to a and paper production is a universal spillovers. measure of impact, but the role that author reputation plays in the life-cycle of the citation rate remains poorly understood. As a result, models of citation dynamics and career trajectories overlook the collaboration and reputation spillovers constitute a cumulative advantage We analyze a large longitudinal career dataset coveringunderlying 350 the competitive aspects of science. To better understand the reputation effect in science, we analyze leading scientists comprising 83,693 papers and 7,577,084the longitudinal ci- citation dynamics of 350 leading scientists from biology, physics, and mathematics. We uncover tations tracked over 384,407 paper years. statistical regularities in the evolutionB. of Longitudinalproductivity and citation impact which dynamics we use as benchmarks for a theoretical model of career growth that we test and validate on real careers. Our model incorporates the life-cycle effect for individual papers, the cumulative advantage arising from scientific reputation, and the preferential attachment effect for citation dynamics.Paper We quality find that is the universally author reputation measured effect dominates according in the to initial the cu- phase of the citation τ II. RESULTS life-cycle, but that themulative preferential number attachment of citations mechanismc(τ emerges)= ast=1 the∆ mainc(t component), where we behind the sustained citation rate of highlydefine cited papers.∆c(t) Comparingas the number between of the citations three disciplines, received we show by the that pa- the impact life-cycle differs between fields:per the in axiomatic year t where discoveriesτ = int mathematicst +1defines have a very the long relation shelf-life, be- whereas the rapid A. Longitudinal productivity dynamicspace in biology and physics results in a shorter half-life despite0 the intense citation rate in the field. tween the paper age τ, the career− age t, and the first year the We model the career trajectory as a sequence of scientific paper was cited, t0. The total number of citations to the pa- Todo: by construction (L corresponds to the career length of in- outputs which arrive at the variable rate n (t). Since the rep- pers coauthored by individual i is calculated by summingi over i c (r, t), the rank-ordered citationdividual distribution,i at the giving timeC of(t data)= extraction). Fig. 1 shows utation of a scientist is typically a cumulativePerform representation statistical χ2 significancei tests on the DGBD • N(t) the characteristic production trajectory obtained by averag- Growthc(r, patterns t). In in order “superstar” to extract academic the careers characteristic scaling tra- of his/her contributions, we consider the cumulativeprofiles for produc- datasets [D] andr [E]=1 and put in SI. ing together the A individual trajectories N˜i(t) belonging to t _ jectory⟩ 102 of C(t), we factorCollaborationCumulative out the reputation scale measures:ci which canA vary tion Ni(t) ni(t) as a proxy for career achieve- (t) ! c ˜ t=1 ! networkeach dataset, N(t) N (t)/A n , where we define N _ i i π ρ τ ⟨ [A/B] 1.30(1) 1.5 i=1 ≡ Calculate i, i, and i for each, of 350 scientists, put(normalized) in α ment. In order to analyze the average properties of Ni(t) considerably[C] 1.15(2) across scientists, and average˜ the resulting≡˜ tra- • [D] 1.55(1) publication N (t) = N(t) / N(1) . tables, and look for relations to[E] other 1.01(1)˜ factors trajectory ~ t jectories1 C (t) C (t)/ c for the A scientists constituting for all the scientists in our sample, we define the normal- 10 i i i This regularity reflects the abundance of of careers with ˜ ˜≡ 1 A ized trajectory Ni(t) Ni(t)/ ni . The quantity ni is each dataset, C(t) (normalized)i=1 Ci(t)/A ci_. Fig. 1 demon- a αi > 1 correspondingζ to accelerated career growth. This ac- ≡ Publication trajectory 0 ≡ cumulative ˜ 10 0 1 ˜ ˜ the average annual production of author i, with Ni(Li)=Li strates10 the_ super-linear10 scalingcitation celerationC(t) ~= t isC consistent(t) / C(1) with increasing returns arising from 4 ⟩ 10 " 3 trajectoryi ∼ (t) I. INTRODUCTION! knowledge and production spillovers. C [A/B] 2.52(1) ⟨ 3 , 10 [C] 2.42(4) ζ > α >1 ⇒ super-linear growth [D] 2.65(1) [E] 1.39(3) We analyze a large longitudinal10 career2 dataset covering 350 1 leading scientists comprising 83,69310 papers and 7,577,0841 ci-Cumulative advantage ~ Citation trajectory b careers become “attractors” of new 100 B. Longitudinal citation dynamics 100 101 opportunities instead of “pursuers” tations tracked over 384,407 paper years. career age, t Citation network

Paper quality is universally measured according to the cu- τ II. RESULTS mulative number of citations c(τ)= t=1 ∆c(t), where we define ∆c(t) as the number of citations received by the pa- A. Longitudinal productivity dynamics per in year t where τ = t t0 +1defines the relation between the paper age τ, the career− age t, and the first year the We model the career trajectory as a sequence of scientific paper was cited, t0. The total number of citations to the papers coauthored by individual i is calculated by summing over outputs which arrive at the variable rate ni(t). Since the rep- ci(r, t), the rank-ordered citation distribution, giving C(t)= utation of a scientist is typically a cumulative representation N(t) of his/her contributions, we consider the cumulative produc- r=1 c(r, t). In order to extract the characteristic scaling tra- t tion Ni(t) ni(t) as a proxy for career achieve- jectory of C(t), we factor out the scale ci which can vary t=1 ment. In order≡ to analyze the average properties of N (t) considerably across scientists, and average the resulting tra- i ˜ for all the scientists in our sample, we define the normal- jectories Ci(t) Ci(t)/ ci for the A scientists constituting ≡ A ized trajectory N˜ (t) N (t)/ n . The quantity n is each dataset, C˜(t) Ci(t)/A ci . Fig. 1 demon- i ≡ i i i ≡ i=1 the average annual production of author i, with N˜ (L )=L strates the super-linear scaling C(t) = C˜(t) / C˜(1) i i i ∼ What makes science special (complex)?

5 _ Interactions mediated by social “forces”: ⟩ 102 Collaboration

(t) ! c ! network

N Collaboration (attractive) ⟨ [A/B] 1.30(1) 1.5

, • [C] 1.15(2) [D] 1.55(1) Competition for priority (repulsive) [E] 1.01(1) • 101 1 • Knowledge (an “exchange particle”) a Publication trajectory 0 10 0 1 10 _ 10 4 ⟩ 10 " 3 i (t) !

C [A/B] 2.52(1) ⟨ 3 , 10 [C] 2.42(4) principal [D] 2.65(1) investigator 102 [E] 1.39(3)

101 1

Citation trajectory b 100 100 101 career age, t Citation network

FIG. 1: Longitudinal analysis of publication and citation growth patterns. (a,b) Growth curves, appropriately rescaled to start from unity, show the characteristic career trajectories of the scientists in each cohort. The characteristic α and ζ exponents shown in each legend are calculated over the growth phase of the career, in (a) over the first 30 years and in (b) over the first 20 years. The mathematicians [E] have distinct career trajectories, with α 1 since collaboration spillovers play a smaller role in their production growth. (c) Schematic illustration of the multiplex scientific network≈ surrounding career i. Links in the upper network represent the dynamic collaborations between scientists (nodes); links within the lower network represent the citation network between papers (nodes); the cross-links between the networks represent the association between individual careers and the corresponding publication portfolio, serving as a platform for reputation signaling [14, 21, 23]. Diverse collaboration strategies

5 _ Interactions mediated by social “forces”: ⟩ 102 Collaboration

(t) ! c ! network

N Collaboration (attractive) ⟨ [A/B] 1.30(1) 1.5

, • [C] 1.15(2) [D] 1.55(1) Competition for priority (repulsive) [E] 1.01(1) • 101 1 ⤷ • Knowledge (an “exchange particle”) a 451 Publication trajectory 0 publications 10 0 1 Watson-Crick strategy: 10 _ 10 4 * Michael Stuart Brown ⟩ 10 " 3 i (t) * Joseph L. Goldstein !

C [A/B] 2.52(1) ⟨ 3 , 10 [C] 2.42(4) Recipients of the 1985 Nobel Prize in Physiology or [D] 2.65(1) diverse collaboration Medicine for describing the regulation of cholesterol 434 metabolism. (95%) 102 [E] 1.39(3) strategies even within the same ﬁeld! 101 Solo-artist strategy: 1 * Marilyn Kozak Citation trajectory b 458 100 N = 70, Nsolo = 59 (84%) 100 101 publications career age, t Citation﹛ network

5 _ ⟩ 102 Collaboration

(t) ! c ! network

N Complexity ⟨ [A/B] 1.30(1) 1.5 , [C] 1.15(2) [D] 1.55(1) [E] 1.01(1) 101 • coevolutionary system 1 • behavioral components a • embedded social processes Publication trajectory 0 10 0 1 10 _ 10 reputation 4 • ⟩ 10 " 3 i (t) ! • economic incentives C [A/B] 2.52(1) ⟨ 3 , 10 [C] 2.42(4) [D] 2.65(1) 102 [E] 1.39(3) Reputation and Impact in Academic Careers, ArXiv: 1303.7274 A. M. Petersen, S. Fortunato, R. K. Pan, K. Kaski, O. Penner, M. Riccaboni, H. E. Stanley, F. Pammolli 101 1

Citation trajectory b 100 100 101 career age, t Citation network

Web of Knowledge. We restrict our analysis to publications de- Acknowledgments AMP acknowledges support from noted as “Articles”, which excludes reviews, letters to editor, COST Action MP0801 “Physics of Competition and Con- corrections, etc. For a given journal j we aggregate papers to- flicts.” AMP, MR, and FP acknowledge support from PNR gether and create a registry of surname and first/middle-initial project “CRISIS Lab.” pairs Surname,Competitive FM . For each journal datasetarenas we select the set of{ surnames which} appear with only one unique FM in the entire database. We assume that there is no intrinsic bias associated with surname, and hence, the set of “rare” surname profiles is a representative sample from the entire distribution of careers [22]. For each pair Surname, FM we then aggregate publication, coauthorship,{ and citation} totals which measure the career achievement of a given author i in a given journal j. Disambiguation strategy: Analyzed subset of “rare” surnames Journal Years Articles Authors, N j CELL 1974-2012 12,349 19,491 (1,753) Nat./PNAS/Sci. 1958-2012 219,656 112,777 (14,478) NEJM 1958-2012 18,347 33,149 (2,897) PRL 1958-2012 98,739 55,827 (10,206)

TABLE I: Summary of journal datsets. N j is the number of unique surnames we were able to identify in each journal j over the denoted period. The N j value in parentheses denotes the number of careers with Li 5. ≥

[1] Stephan, P. 2012 How Economics Shapes Science, Harvard Uni- [12] Wutchy, S., Jones, B. F., Uzzi, B. 2008 The increasing domi- versity Press, Cambridge MA, USA. nance of teams in production of knowledge. Science 322, 1259– [2] Petersen, A. M., Riccaboni, M., Stanley, H. E., Pammolli, F. 1262. 2012 Persistence and uncertainty in the academic career. Proc. [13] Borner,¨ K., Contractor, N., Falk-Krzesinski, H. J, Fiore, S. M., Nat. Acad. Sci. USA 109, 5213 – 5218. Hall, K. L., Keyton, J., Spring, B., Stokols, D., Trochim, W., [3] Merton, R. K. 1968 The Matthew effect in science. Science 159, Uzzi, B. 2010 A multi-level systems perspective for the science 56–63. of team science. Sci. Transl. Med. 2, 49cm24. [4] Petersen, A. M., Wang, F., Stanley, H. E. 2010 Methods for [14] Palla, G., Barabasi,´ A. L., Vicsek, T. 2007 Quantifying social measuring the citations and productivity of scientists across group evolution. Nature 446, 664–667. time and discipline. Phys. Rev. E 81, 036114. [15] Chait, R. P., ed. 2002 The Questions of Tenure, Harvard Univer- [5] Petersen, A. M., Jung, W.-S., Yang, J.-S., Stanley, H. E. 2011 sity Press, Cambridge MA, USA. Quantitative and empirical demonstration of the Matthew effect [16] Segalla, M., Rouzies,´ D., Flory, M. 2001 Culture and career ad- in a study of career longevity. Proc. Natl. Acad. Sci. USA 108, vancement in Europe: Promoting team players vs fast trackers. 18–23. European Management Journal 19, 44–57. [6] Petersen, A. M., Fortunato, S., Pan, R. K., Kaski, K., Pen- [17] Kaplan, K. 2010 The changing face of tenure. Nature 468, 123– ner, O., Riccaboni, M., Stanley, H. E., Pammolli, F. 2013 125. Reputation and impact in academic careers. Submitted, e-print: [18] Powell, K. 2012 Off the tenured track. Nature 491, 627–629. arXiv:1303.7274 [19] Kaminski, D. & Geisler, C. 2012 Survival Analysis of Faculty [7] Jones, B. F., Wuchty, S., Uzzi, B. 2008 Multi-University Re- Retention in Science and Engineering by Gender. Science 335, search Teams: Shifting Impact, Geography, and Stratification 864–866. in Science. Science 322, 1259–1262. [20] Hirsch, J. E. 2008 Does the h index have predictive power. Proc. [8] Cole, J.R. 1981 Social Stratification in Science, Chicago, Illi- Natl. Acad. Sci. 104, 19193–19198. nois, The University of Chicago Press. [21] Acuna, D. E., Allesina, S., Kording, K. P. 2012 Future impact: [9] Franzoni, C., Scellato, G., Stephan, P. 2011 Changing Incen- Predicting scientific success. Nature 489, 201–202. tives to Publish. Science 333, 702–703. [22] Mazloumian, A. 2012 Predicting Scholars’ Scientific Impact. [10] Borner,¨ K., Maru, J. T., & Goldstone, R. L. 2004 The simultane- PLoS ONE 7. ous evolution of author and paper networks. Proc. Natl. Acad. [23] Penner, O., Petersen, A. M., Pan, R. K., Fortunato, S. 2013 The Sci. USA 101, 5266–5273. case for caution in predicting scientists’ future impact. Phys. [11] Guimera, R., Uzzi, B., Spiro, J., Amaral, L. A. N. 2005 Team Today 66, 8–9; O. Penner, R. K. Pan, A. M. Petersen, S. Fortu- assembly mechanisms determine collaboration network struc- nato. Vetting career predicability models. Submitted (2013). ture and team performance. Science 308, 697–702. [24] Fu, D., Pammolli, F., Buldyrev, S. V., Riccaboni, M., Matia, 5

tire dataset. This strategy was recently employed by [22] to analyze the career predictability problem. We confirm that the results reported in [4, 5], which analyzed this high-impact career dataset without using a pruning method, remain unchanged. Hence, we validate the assumption that analyzing relatively small subsets of careers within a specific journal can significantly reduce the problems arising from the name disambiguation problem. Using this name pruning method, Fig. 2 shows the number of “unique” authors per year, a sample which accounts for roughly 20-25% of the total names in the dataset (see Section IV B for more detail), and hence grows at the same rate as the total indistinct number of coauthors taken from all papers in a given year. We observe the same pattern for the other journals which we also analyzed separately as a robustness check. Ta- FIG. 4: Proportional growth model for career production. (a) We ble I lists the raw number of careers analyzed, which for the aggregate all ri(t) values for each discipline into a single datasetPeering and triad inside of PNAS/Nature/Science the high-impact resulted arena in 112,777 “unique” calculate the empirical probability density function (pdf) P (r). The careers in those journals alone, 14,478 of which had a dura- maximum likelihood estimation of P (r) for the double-exponential tion of 5 or more years between their first and last publication. (Laplace) distribution (grey curve) shows good agreement. Interest- ingly, the distribution for physics and biology are characterizedNature/PNAS/Science by Analyzing the distributionBasic measures of career for measures survival in a particular 100 approximately equal width (σ), whereas the distribution forLc math= 1 (100%) journal,a we verify the followingand achievement stylized facts for all journals P(L =1) = 67% is significantly more narrow, highlighting the10 importance1 of disci- analyzed. First, the length of time between the first and last Lc = 5 (23%) j j j plinary context in evaluating career profiles. (b) To test the stability publication ofLongevity author i, L t t +1, in a given high- 102 Lc = 10 i ≡ i,f − i,0 of the distribution over career trajectory subintervals, we separate impact(14%) journalin ja, given is extremely journal set right-skewed, is extremely right- as illustrated in P(L) ri(t) values into 5 non-overlapping 10-year periods103 and verify the Fig. 5(a). Mostskewed, careers in agreement enter and with exit the with quantitative the same publica- stability of the Laplace P (r) for the careers in dataset4 [D]. For each j predictions of a rich-get-richer career progress 10 tion, i.e. L =1. However, the champions of these “compet- P (r), we also plot the corresponding Laplace distribution (solid line) i model using the maximum likelihood estimator method. To10 improve0 graph- 101 itive arenas” continue to publish for roughy their entire sci- ical clarity, we vertically offset each P (r) by a constant factor.longevity, (c) Lentific lifetime. This statistical regularity was also shown to Accounting for individual production factors by using the standard- hold in professional sports (baseball, basketball, and football) 100 r ) P (r) as predictedb by a position-dependent Matthew effect model ized production change , the resulting pdfsp collapse onto 101 a Gaussian distribution with zero mean and unit N variance. (d) The for career growth [5].

! 2 cumulative distribution CDF(X Si) for each10 discipline is ap- ≥ Using the setLikewise, of authors since with productionLi is5 ,highly we plot correlated in Fig. 5(b) proximately exponential for small S, The exponential103 distribution is ≥ the cumulativewith distribution longevity, ofthe total distribution number of of cumulative publications Np a key requirement for showing that the unconditional4 distributions-2 10 Np which is alsopublications extremely right-skewed.is also extremely Although right-skewed it is not the P (r) in (a) and (b) follow from an exponentialCDF (x mixing of conditional 105 purpose of this(Lotka’s analysis law) to specify precisely the functional Gaussian distributions P (r Si) [2]. As a visual comparison0 guide,1 2 | 10 10 form of10 these distributions, it is notable that the probability we plot the dashed line representing an exponential distributiontotal publications, with N mean S =14coauthors. distributionp is approximately Pareto (also known in this con- 100 3 text as Lotka’sc law) [28, 29], P (Np) 1/Np , but with some 1 ∼ 10 clear curvature reflecting finite-size effects arising from im- cant substitution in risk, since online visibility102 is a new and portant limitations such as human longevity, incomplete ca- ~ growing competitive arena in science. 103 reers, etc. P(C) In this section we investigate the distribution104 of longevity,Log-normal fit Analyzing the cumulative impact of each author’s publi- productivity, and impact in these 3 journals10 which5 are widely cations within this journal triad is complicated by the fact 2 1 0 1 2 regarded as the elite multidisciplinary journals.10 Since the10 pub-10 that10 citations10~ are time-dependent as well as discipline depen- lications in high-impact journals are those whichtotal constitute normalized a citations,dent. However, C by calculating a standardized citation measure significant portion of a scientist’s reputation, we analyze the which discounts the total citation count by the average number partial but significant career profile of the scientists within of citations for all papers published in the same year one can these journals. Furthermore, since the journals chosen are approximately remove the underlying time dependence and relatively select, and hence, relatively small as compared to achieve universal log-normal citation distributions [30, 31]. the entire set of scientific publications, we are able reduce the Hence, we use the normalized impact transformation false positive scenario in which two or more careers are joined and analyzed as one (the name disambiguation problem). c˜ = cj (t)/ cj(t) (6) p In order to further overcome the disambiguation problem j of distinguishing between multiple authors with the same sur- where cp(t) are the total number of citations observed at name and first name abbreviation, we select the set of indi- present time T for paper p from journal j in year t and cj(t) viduals with “unique” names corresponding to the occurrence is the average over all papers from the same year. Using this of only one first and middle initial for that surname in the en- impact measure it is then possible to simply aggregate across “Cumulative advantage” For each career i we track his/her longitudinal publication rate by aggregating over publications in a specific set of high-impact journals

(1) (2) (3) (4) (n) career i 1 1 2 3 4 5 n t Q: What is the characteristic waiting time τi(n) between an author’s nth paper and (n+1)th paper? b H. Eugene Stanley

impact 103 “Cumulative advantage”

and For each career i we track his/her longitudinal publication rate by aggregating 2 10 over publications in a speciﬁc set of high-impact journals output 101 (1) (2) (3) (4) (n)

0 Scientiﬁc 10 career i 0 10 20 30 40 1 1 2 3 4 5 n t 4 careers with Q: What is the characteristic L ≥ 5 and Np ≥ 10 Nature/PNAS/Science 3 CELL waiting time τi(n) between an NEJM author’s nth paper PRL th

years and (n+1) paper?

, 2 " ) n

( By the 10th paper, the waiting ! 1 ! time between publications has decreased by ~ factor of 2! 0 0 10 20 30 40 paperpaper n n PETERSEN, WANG, AND STANLEY PHYSICAL REVIEW E 81, 036114 ͑2010͒

TABLE IV. Summary of paper shares for “completed” careers. institutions which produce over many years many significant The value of the log-normal fit parameters ␮ and ␴ correspond to results per year. Furthermore, the organization of the curves c the pdf before the cutoff value of Ps Ϸ2 paper shares. The values of in Fig. 7 suggests that it is more difficult at the beginning of c ␣ are calculated using a data values after the cutoff Ps ϵ1 paper a career to repeatedly publish in CELL than PRL. Reaching a shares, which corresponds to approximately 8% of the total data for crossover point along the career ladder is a generic phenom- each journal. enon observed in many professions. Accordingly, surmount- ing this abstract crossover is motivated by significant per- Journal ␮ ␴ ␣ sonal incentives, such as salary increase, job security, and CELL −1.7Ϯ0.1 0.7Ϯ0.1 2.60Ϯ0.05 managerial responsibility. NEJM −1.7Ϯ0.1 1.0Ϯ0.1 2.60Ϯ0.02 Nature −1.3Ϯ0.1 1.0Ϯ0.1 2.74Ϯ0.05 IV. DISCUSSION PNAS −1.6Ϯ0.1 0.7Ϯ0.1 2.56Ϯ0.02 Scientific careers share many qualities with other com- PRL −1.1Ϯ0.1 1.0Ϯ0.1 2.35Ϯ0.02 petitive careers, such as the careers of professional sports Science −1.4Ϯ0.1 0.9Ϯ0.1 2.61Ϯ0.02 players, inventors, entertainers, actors, and musicians ͓15,32,33͔. Limited resources such as employment, salary, creativity, equipment, events, data samples, and even indi- n and the paper n+1. The values of ͗␶͑1͒͘ for each journal vidual lifetime contribute to the formation of generic arenas are 2.2 ͑CELL, PRL͒, 3.0 ͑Nature, PNAS, Science͒ and 3.5 for competition. Hence, of interest here is the distribution of ͑NEJMQ͒uyears.anThetidecreasetatiinvwaitinge atimendbetween empubli-pirsuccessicaandl dproductivityemoin nhigh-impactstrajournalstionwhich oinf the Matthew effect in a cations is a signature of the cumulative advantage mecha- principle have high standards of excellence. nism qualitatively described in ͓19͔ and quantitatively ana-stuIndscience,y otheref careaunwrittenreeguidesr loto successngerequiringvity lyzed in ͓16,18͔. To avoid presentingModelingstatistical fluctuations the ingenuity“Rich-get-richer”, longevity, and publication. We observe effecta quantifi- arising from theAsmalllexsizeandeof datar sets,M. wePeonlytepresentrsen1,able Woo-Sungstatistical regularity Jdescribingung1,publication Jae-Sukcareers Yof ang2, H. Eugene Stanley1 ͗␶͑n͒͘ computed for data sets exceeding 75 observations. 1 individual scientists across both time and discipline. Interest- To explain the steady decline of Ctheencurveter foforr PPRLolywemer inglyStu,dwiee sfind anthatd Dtheescalingpartmexponentent of forPhindividualysics, Bpapersoston University, MA, USA mention thatForwardPRL has many progressauthors 2withDep manyafollowsrtmarticlesent o f aP ͑hstochastic␥yϷs3ic͒ iss,lar Kgerorethana “progress Uthenivscalingersitexponenty, S rate”eouforl 1total3 g(x)6-citation701 , Republic of Korea ͑nտ100•͒. A possible explanation is that a significant number shares ͑␣Ϸ2.5͒ and the scaling exponent for total paper of these authors are involved in large particle accelerator shares ͑␣Ϸ2.6͒, which indicates that there is a higher fre- experiments with multiple collaborating groups. TheseSmul-tochaquencystic mofostellardel fcareersor cathanreestellarr propapers.gressThis: spisaconsis-tial Poisson process AtilateralbstraprojectsctCumulativecontribute significantly advantage:to the heavy tail g(x)ob- increasestent with the observation withthat careera stellar career positioncan result from x Decreasing inter-publication time !(n) The Matthew effect refers to the adage written s•ome two-thousand years ago in the Gospel of St. Matthew: ``For to all those who have, mservedore will bine gtheivenpdf". Eveofn tthewo mnumberillennia laofterauthors, this idiomper is paper ͑Fig. 3͒. an arbitrary combination of stellar papers andgconsistent(x) suc- used by sociologists to qualitatively descHence,ribe indivtheidual decayprogressin anthed thecurve interplafory betPRLween swhichtatus approaches ... and reward. Quantitative studies of professional careers are traditionally limited by the difficulty in cess, as demonstrated in Table III. In all, the statistical regu- measuring progress and the lack of datazero on inmightdividualbe cardueeers.to Hothewevprojecter, in somleaderse profesatsionlars,ge experimental larity found in the distributionsg(5) for both citation shares and there are well-defined metrics that quantify career longevity, success, and prowess, which together g(3) g(4) contribute to the overall success rating for an individual employee. Here we demonstrate testable paperg(2)shares lend naturally to methods based on extreme evidence, inherent in the remarkable statistical reg1ularity of career longevity distributions, of the g(1) statistics in order to distinguishing stellar careers. Such ‹ !(n) › = 1 / g(n) age-old Matthew ``rich get richer" effect, in which longevity and past success lead to cumulative advantage. We develop an exactly solvable stochastic model that quantitatively incorporates the methods have been developed for Hall of Fame candidacy in

Matthew effect such that it can be validated in co> 0.8mpetitive professions. These results demonstrate baseball ͓16,34͔, where statistical benchmarks are estab- that statistical laws can exist at even the microscopic social level, where the collective behaviour of individuals can lead to emergent phenomena. We test our model on the careers of 400,000 lished using the distribution of success. (1) career

τ career position, x scientists using data from six high-impact journals0.6. We further confirm our findings by testing the Statistical physicists have long been interested in complex

model on the careers of more than 20,000 athlet< es in four sports leagues.

/ CELL interacting systems, and are beginning to succeed in describ-position, x > 0.4 NEJM ing social dynamics using models thatx-1were developedx x+1in the ) 1 2 3 4 5 Nature We analyze the inter-publication waiting time between an author’s paper n and n !(n)

( PNAS context of concrete physical systems ͓35͔. This study is in- “One-hit wonders” andτ “IronPRL Horses” We model progress in competitive professions as a random hopping process with two main ingredients: paper n+1 in a given journal. The quantity !(n) is inversely proportional to the . • random forwaspiredrd progrebyss uthep the longcareer termladder goal of using quantitative methods < 0.2 Science . progress probability g(x) used in the stochastic model. We find that the average • random stopping times, terminating the career from statistical physics to answer traditional questions rooted inter-publication time ‹ !(n) › decreases with increasing number publications, Completed Careers 1958-2008 We solve the correspondingg(x) master equa tio=n gov er1ning th/e ev⟨oluτtion(x) of P(x,t),⟩ the probability that an individual is at career . in social science ͓36͔, such as the nature of competition, consistent with the Matthew effect. The values of ‹ !(1) › are 2.2 (CELL, PRE), 3.0 0 position x at time t. The progress rate parameter g(x) determines the relative difference in late-career progress versus 0 1 2 (Nature, PNAS, Science), and 3.5 (NEJM) years. 10 10 10early-career prosuccess,gress. We cproductivityhoose a functio,naandl formthe for guniversal(x) that increfeaturesases with x,of caphumanturing the salient feature of the Author’s nth paper Matthew effect activitythat it bec.oManymes easistudieser to makebegin progresass thempiricale further alondescriptions,g is the career. such See Ref. [3] for more details. The progress probability g is the See Ref. [1] for more details. Lou Gehrig HOF plaque as the studies of common mobility patterns ͓37͔, sexuality FIG. 7. Color online A decreasing waiting time ␶ n between ͑ ͒ ͑ ͒ ͓38,39͔, and financial fluctuations ͓40͔, and lead to a better publications in a given journal suggests that a longer publication understanding of the underlying mechanics. It is possible that career ͑larger n͒ facilitates future publications, as predicted by the inverse of the mean waiting time τ Career success metrics in sports theTempiricalhe “Rilawsch-reportedget-riherechewillr”motivate Matthusefulew descrip-effect Matthew effect. We plot ͗␶͑n͒͘/͗␶͑1͒͘, the average waiting time Methods for measuring the citations and tive theoriesQuantitativeof success and empiricaland productivity demonstrationin competitiveof the en- ͗␶͑n͒͘ between paper n and paper n+1, rescaled by the average - ``For to all those who have, more will be given” productivity of scientists across time and Matthew effect in a study of career longevity. A. M. waiting time between the first and second publication, ͗␶͑1͒͘. The vironments. Matthew 25:29 discipline, A. M. Petersen, F. Wang, H. E. Petersen, W.-S. Jung, J.-S. Yang, H. E. Stanley. Proc. values of ͗␶͑1͒͘ are 2.2 ͑CELL, PRL͒, 3.0 ͑Nature, PNAS, Science͒, 3 Stanley. Phys. Rev. E 81, 036114 (2010). Natl. Acad.Xc Sci.# 1 USA0 108, 18-23 (2011). and 3.5 ͑NEJM͒ years. Physical Review Letters exhibits a more ACKNOWLEDGMENTS rapid decline in ␶͑n͒, reflecting the rapidity of successive publications ͑often by large high-energy experiment collaborations͒, which We thank S. Miyazima and S. Redner for helpful com- is possible in this high-impact letters journal. ments and the NSF for financial support. For " > 1 : P(x) is bimodal In sports, successes are obtained in proportion to the total number of opportunities. " = 0.40 036114-8 Hence, the probability density function P(z) of career successes z is also a truncated For " < 1 : P(x) is a truncated power-law, power law with the same scaling exponent " as the corresponding longevity xw distribution.

Hits RBI • (A) MLB Baseball: xc ! 1200, xc ! 600. One hit wonders: 5% of all fielders 1920-Present finish career with only 1 hit ! 3% of all pitchers finish career with less than an inning pitched! Points Rebounds • (B) NBA Basketball: xc ! 8000, xc ! 3500 We choose a functional form for the progress rate g(x) which is characterized by two parameters: Furthermore, we approximate P(z) with the Gamma pdf, and use the extreme statistics of the Gamma distribution to estimate benchmarks which distinguish stellar We analyze the professional careers of: careers (e.g. Hall of Fame). See [1] and [2] for a discussion of establishing statistically • 400,000 scientists publishing in 6 high-impact journals: Nature, the Proceedings of the " significant milestones for HR, K, RBI, and W in professional baseball. (1) " is a scaling exponent which quantifies the growth of g(x) for small values of x. For small x < xc the progress rate g(x) ~ x See Ref. [1,2,4] for more details. National Academy of Science, Science, CELL, the New England Journal of Medicine, and Physical Review Letters Two different types of career longevity probability density function (pdf) emerge depending on the value of " : (i) For convex > 1 it is more difficult to make progress early in the career, and hence, P(x) is bimodal, with one group of stunted • 20,000 professional athletes: Major League Baseball (1920-2004), Korean " References careers grouped around small x < x values and another group of successful careers grouped around larger x > x values. Professional Baseball League 1982-2007, National Basketball Association 1946-2004, c c [1] A.M. Petersen, W.-S. Jung, J.-S. Yang, H.E. Stanley. Quantitative and English Premier League 1992-2007 empirical demonstration of the Matthew effect in a study of career (ii) For concave " < 1 it is easier to make progress early in the career. This feature results in a remarkable statistical regularity longevity. Submitted. arXiv:0806.1224 Theoretical curves (solid green lines) derived from our stochastic model show excellent over several orders of magnitude captured by a truncated power-law with scaling exponent ". agreement with empirical data. We define metrics for career longevity that are inherently [2] A.M. Petersen, O. Penner, H.E. Stanley. Detrending career statistics in related to the time spent in the career, and according to the available data. professional baseball: Accounting for the steroids era and beyond. • scientific longevity: x = y - y +1 last first (2) x is a career length scale which separates newcomers from veterans on the career ladder. The width x of the “potential barrier” arXiv:1003.0134 which is the time interval in years between a scientist’s first and last publication in a c w -1/" given high-impact journal which newcomers must overcome scales as xw / xc ! xc [3] A.M. Petersen, F. Wang, H.E. Stanley. Methods for measuring the citations and productivity of scientists across time and discipline. Physical • sports longevity: x = total number of in-game opportunities over the career We observe " < 1 for all careers analyzed. The statistical regularity implies that the relative number of individuals with career longevity Review E 81, 1 (2010). arXiv:0911.1322 x and x are given approximately by the ratio P(x )/P(x ) = (x /x )" which is quantified only by a scale-free ratio and the scaling - Baseball: At-bats (AB), Innings Pitched in Outs (IPO) 1 2 1 2 2 1 [4] A.M. Petersen, W.-S. Jung, H.E. Stanley. On the distribution of career - Basketball: minutes played exponent. longevity and the evolution of home run prowess in professional baseball. - Soccer: games played See Ref. [1] for more details. See Ref. [1,2,3,4] for more details. Europhysics Letters 83, 50010 (2008). Statistical regularities in the career longevity distribution 0 1 2 3 4 1 2 3 4 5 10 10 10 10 10 10 10 10 10 10 Major League Baseball -2 A B -2 10 α = 0.77 10 α = 0.63 130+ years of player -4 • -4 10 10 Baseball (fielders) Basketball (NBA) statistics, ~ 15,000 careers AB (Korea, KPB) -6 -6 Minutes 10 10 AB (USA, MLB) -1 ``One-hit wonders” 10

P(x) C D P(x) α = 0.71 α = 0.55 3% of all ﬁelders ﬁnish their -2 • -3 10 10 career with ONE at-bat! Pro Sports

-5 Baseball (pitchers) -4 10 Soccer (Premier League) 10 3% of all pitchers ﬁnish their IPO (Korea, KPB) • IPO (USA, MLB) Games career with less than one -7 -6 10 0 1 2 3 4 0 1 2 3 10 10 10 10 10 10 10 10 10 10 inning pitched! x (career longevity in opportunities) 0 0 10 10 ``Iron horses” -1 α ≈ 0.40 E F -1 10 10

-2 -2 10 10 • Lou Gehrig (the Iron Horse): NY

P(x) -3 -3 P(x) 10 Nature CELL 10 Yankees (1923-1939) PNAS NEJM -4 -4 10 Science PRL 10 • Played in 2,130 consecutive games in -5 -5 15 seasons! 8001 career at-bats! 10 0 1 0 1 210 Academia 10 10 10 10 10 Career & life stunted by the fatal x (career longevity in years) • neuromuscular disease, amyotrophic opportunities ~ time duration lateral sclerosis (ALS), aka Lou Gehrig’s Disease 7

2) We run the Monte Carlo (MC) simulation for T 100 time periods and all agents are by construction from the same age cohort (born at same time). ≡

3) Each time period corresponds to the allocation of P I n opportunities, sequentially one at a time, to ≡ i=1 0,i randomly assigned agents i,wheren0,i 1 is the potential production capacity of a given individual. ≡ 4) The assignment of a given opportunity is proportional to the time-dependent weight (capture rate) wi(t) of each agent. Hence, the assignment of 1 opportunity to agent i at period t results in the production (achievement) n (t) to increase by one unit: n (t) n (t) + 1. In the next time period t + 1, we update the weight w (t + 1) i i → i i to include the performance ni(t)inthecurrentperiod.

B. Initial Condition

The initial weight at the beginning of the simulation is w (t = 0) n for each agent i with n 1. The value i ≡ c c ≡ nc > 0 ensures that competitors begin with a non-zero production potential, and corresponds to a homogenous system where all agents begin with the same production capacity. Hence, we do not analyze the more complicated model wherein external factors (i.e. collaboration factors) can result in a heterogeneous production capacity across scientists. By construction, each agent begins with one unit of achievement n (t = 1) 1. i ≡ Modeling competition Agent-based model of competition with C. System Dynamics achievement appraisal

1) In each Monte Carlo step we allocate one opportunity to a randomly chosenAchievement individual measuredi so that by ni(t), the nnumberi(t)+1 of opportunities (ex. publications) captured in time period→ t 2) The individual i is chosen with probability (t) proportional to [w (t)]π Pi i π I = ﬁnite labor wi(t) i(t)= (S16) P I π force size i=1 wi(t) where the value wi(t) is given by an exponentially weighted sum over the entire achievement history t 1 − c∆t w (t) n (t ∆t)e− . (S17) i ≡ i − ∆t=1 Persistence and Uncertainty in the Academic Career, A. M. Petersen, M. Riccaboni, H. E. Stanley, F. Pammolli. Proc. Natl. Acad. Sci. USA 109, 5213-5218 (2012). The parameter c 0 is a memory parameter which determines how the record of accomplishments in the past aﬀect the ability to≥ obtain new opportunities in the current period, and therefore, the future. The limit c =0 rewards long-term accomplishment by equally weighting the entire history of accomplishments. Conversely, when c 1 the value of wi(t) is largely dominated by the performance ni(t 1) in the previous period, corresponding to increased emphasis on short-term accomplishment in the immediate− past. Intermediate values 0 1, and sub-linear capture π < 1.

4) At the end of each time period, the weight wi(t) is recalculated and used for the entirety of the next MC time period corresponding to the allocation of the next I n achievement opportunities. × c

D. Model Results

We simulate this system for a realistic labor force size I = 1000 with the assumption that in any given period, an individual has the capacity for one unit of production (n 1). We evolve the system for T = 100 periods c ≡ corresponding to I nc T Monte Carlo time steps. The timescale T represents the (production) lifetime of individuals with ﬁnite longevity.× In× this model we do not include exogenous shocks (career hazards) that can result in career death [16]. Here we analyze four quantities:

T 1) The distribution P (N) of the total number of opportunities Ni(T ) t=1 ni(t) captured by agent i over the course of the T period simulation. ≡ − 7

2) We run the Monte Carlo (MC) simulation for T 100 time periods and all agents are by construction from the same age cohort (born at same time). ≡

B. Initial Condition

The initial weight at the beginning of the simulation is w (t = 0) n for each agent i with n 1. The value i ≡ c c ≡ 5 nc > 0 ensures that competitors begin with a non-zero production potential, and corresponds to a homogenous system where all agents begin with the same production capacity. Hence, we do not analyze the more complicated model wherein external factors (i.e. collaboration factors) can result in a heterogeneousD. production A Proportional capacity across growth scientists. model for scientiﬁc Dataset A output By construction, each > agent begins with one unit of achievement ni(t = 1) 1. i Dataset B A ≡

< n 1 10 ! = 0.74(4) We develop a stochastic model as a heuristic tool to C. System Dynamics better understandAppraising the eff ectsprior of achievement long-term versus short- term contracts. In this competition model, opportunities 1) In each Monte Carlo step we allocate one opportunity to a randomly chosen(i.e.Achievement individual new scientific measuredi so that publications)by ni(t), the nnumberi( aret)+1 of captured opportunities according captured in time period t → toπ a general mechanism whereby the capture rate i(t) 2) The individual i is chosen with probability i(t) proportional to [wi(t)] P P dependsThe cohort on of theI agents appraisal compete forwi a( tfixed) of number an individual’s of opportunities record in π wi(t) ofeach achievement period over a overlifespan a of prescribed t = 1... T periods history.. We define the i(t)= I (S16) Ave. annual output, 0 P π appraisal to be an exponentially weighted average over a 10 i=1 wi(t) In each period, the capture rate of a given individual i is calculated by an given individual’s history of production 1 appraisal of the achievement history where the value) 10w (t) is given by an exponentially weighted sum over the entire achievement history r i t 1 ( Dataset A i − c∆t ! Dataset B t 1 B capture wi(t) ni(t ∆t)e− , (8) − c∆t rate ≡ − w (t) n (t ∆t)e− . (S17) { i ≡ i − ∆t=1 ∆t=1 exponential Appraisal which is characterized by the appraisal horizon 1/c.We The parameter c 0 is a memory parameter which determines how thetimescale record of 1/c accomplishments in the pastdiscount factor use the value c = 0 to represent a long-term appraisal affect the ability to≥ obtain new opportunities in the current period, and therefore, the future. The limit c =0 (tenure)c → 0 : appraisal system over and all lifetime a value achievementsc 1 ( to~ tenure represent system) a short- rewards long-term accomplishment by equally weighting the entire history of accomplishments. Conversely, when termc >1 : appraisalappraisal over system. only recent Eachachievements agent (short-termi =1...I contractsimultane- system) c 1 the value of wi(t) is largely dominated by the performance ni(t 1) in the previous period, corresponding to increased emphasis on short-term accomplishment" / 2 = 0.40(3) in the immediate−ously past. attracts Intermediate new opportunities values 0 1,i and sub-lineari captureuntilπ < all1.P opportunities for a given period t are allo-

(t) A: cated. We assume that each agent has the production i ! = 0.68(1)

4) At the end ofn each time period, the weight w (t) is recalculated and used for the entirety of the next MC time B: ! = 0.52(1) i potential of one unit per period, and so the total number period corresponding to the allocation of the next I n achievement opportunities. C: ! = 0.51(2) × c of opportunities allocated per period P is equal to the 1 10 ! number of competing agents, P I. ≡ D. Model Results We use Monte Carlo (MC) simulation to analyze this 2-parameter model over the course of t =1...T sequen- We simulate this system for a realistic labor force size I = 1000 with thetial assumption periods. thatIn each in any production given period, period (representing a timescale on the order of half a human year), a fixed an individual has the capacity for one unit of production (nc 1). We evolve the system for T = 100 periods ≡ number of P production units are captured by the com- corresponding to I nc0 T Monte Carlo time steps. The timescaleC T represents the (production) lifetime of individuals with finite longevity.×10 In× this model we do not include exogenous shocks (careerpeting hazards) agents. that At the can end result of each in career period, we update each death [16]. Here weNumber of publications, analyze four quantities: wi(t) and then proceed to simulate the next preferential 0 1 capture period t + 1. Since i(t) depends on the relative 10 10 T P 1) The distribution P (N) ofNumber the total of number coauthors, of opportunities k (t) Ni(T ) achievementst=1 ni(t) captured of every by agent, agent i theover relative the competitive ad- i ≡ course of the T period simulation. vantage of one individual over another is determined by − the parameter π. In the SI Appendix text we elaborate FIG. 3: Quantitative relations between career growth, career risk, and collaboration efficiency. The fluctuations in produc- in more detail the results of our simulation of synthetic tion reflect the unpredictable horizon of “career shocks” which careers dynamics. We vary π and c for a labor force of size I 1000 and maximum lifetime T 100 periods as can affect the ability of a scientists to access new creative op- ≡ ≡ portunities. (A) Relation between average annual production a representative size and duration of a real labor cohort. ni and collaboration radius Si Med[ki] shows a decreasing Our results are general, and for sufficiently large system marginal output per collaborator≡ as demonstrated by sublin- size, the qualitative features of the results do not depend ear ψ < 1. Interestingly, dataset [A] scientists have on average significantly on the choice of I or T . a larger output-to-input efficiency. (B) The production fluc- The case with π = 0 corresponds to a random capture tuation scale σi(r) is a quantitative measure for uncertainty ψ/2 model that has (i) no appraisal and (ii) no preferential in academic careers, with scaling relation σi(r) Si .(C) capture. Hence, in this null model, opportunities are cap- Over time, there is an increasing returns evident in∼ the annual tured at a Poisson rate λp =1perperiod.Theresults production ni(t) since α > 1. Management, coordination, and of this model (see Fig. S13) shows that almost all ca- training inefficiencies can result in a γ < 1 corresponding to a decreasing marginal return with each additional coauthor in- reers obtain the maximum career length T with a typical put. The significantly larger γ value for dataset [A] scientists career trajectory exponent αi 1. Comparing to simulations with π > 0 and c 0,≈ the null model is similar seems to suggest that managerial abilities related to output ≥ efficiency is a common attribute of top scientists. to a “long-term” appraisal system (c 0) with sublin- ear preferential capture (π < 1). In→ such systems, the Crowding out by “kingpins”

Our theoretical model suggests that short-term appraisal systems: * can amplify the effects of competition and uncertainty making careers more vulnerable to early termination, not necessarily due to lack of individual 1/c talent and persistence, but because of random negative production shocks. * effectively discount the cumulative achievements of the individual.

* may reduce the incentives for a young scientist to Appraisal timescale 1,000invest 2,000 3,000 in 4,000human 5,000 and social capital accumulation. 1,000 2,000 3,000 4,000 5,000 Longevity probability distributions

2,000 4,000 6,000 8,000 10,000 12,000 2,000 4,000 6,000 8,000 10,000 12,000

Fig. 4. MC simulation of the linear preferential capture model (π 1) for varying contract length parametrizedFig. by 4. c.MC Wesimulation plot the probability of the linear distributions preferential for capture model (π 1) for varying contract length parametrized by c. We plot the probability distributions for ¼ ¼ (i) Ni, the total number of opportunities captured by the end period T,(ii) the growth acceleration exponent(αi)i ,(Niii,) the the total single number period of growth opportunities fluctuation capturedri t by the end period T,(ii) the growth acceleration exponent α ,(iii) the single period growth fluctuation r t i ð Þ i i ð Þ including for comparison the Laplace (solid green) and Gaussian (dashed red) best-fit distributions calculatedincluding using the for respective comparison MLE the estimator Laplace,and( (solidiv) green) the and Gaussian (dashed red) best-fit distributions calculated using the respective MLE estimator,and(iv) the career longevity L defined as the time difference between an agent’s first and last captured opportunity. Results for c → 0 systems shows that for a “long-term i career longevity Li defined as the time difference between an agent’s first and last captured opportunity. Results for c → 0 systems shows that for a “long-term appraisal” scenario careers are less vulnerable to low-production phases, and as a result, most agents sustainappraisal production” scenario throughout careers the are career. less vulnerable Conversely, to low-production phases, and as a result, most agents sustain production throughout the career. Conversely, “ ” “ ” results for c ≥ 1 systems show that for a short-term appraisal scenario the labor system is driven by fluctuationsresults that for canc ≥ cause1 systems career showsudden that for death a “short-termfor a appraisal” scenario the labor system is driven by fluctuations that can cause career “sudden death” for a large fraction of the population. In this short-term appraisal model, there are typically a small number of agentslarge who fraction are able of the to capture population. the majority In this short-termof the appraisal model, there are typically a small number of agents who are able to capture the majority of the production opportunities with remarkably accelerating career growth reflected by significantly large αi ≥ 1.Thus,afew“lucky” agents are able to survive the production opportunities with remarkably accelerating career growth reflected by significantly large αi ≥ 1.Thus,afew“lucky” agents are able to survive the initial fluctuations and end up dominating the system. In SI Appendix: and Figs. S12–S16, we further show thatinitial systems fluctuations with increased and end levels up dominating of competition the system. In SI Appendix: and Figs. S12–S16, we further show that systems with increased levels of competition (π > 1) mimic systems with short-term contracts, resulting in productivity “death traps” whereby most careers stagnate and terminate early. (π > 1) mimic systems with short-term contracts, resulting in productivity “death traps” whereby most careers stagnate and terminate early. before reaching age 0.01T, and 25% of the labor population dies under the review processbefore (3, 6). reaching Critics age of tenure0.01T, argue and 25% that of te- the labor population dies under the review process (3, 6). Critics of tenure argue that te- 0 02 before reaching age . T (see SI Appendix: Table S1). Hence, nure places too much financialbefore reaching risk burden age on0.02 theT modern(see SI com- Appendix: Table S1). Hence, nure places too much financial risk burden on the modern competitive research university and diminishes the ability to adapt to in model short contract systems, the longevity, output, and impact in model short contract systems, the longevity, output, and impact petitive research university and diminishes the ability to adapt to of careers are largely determined by fluctuations and not by per- shifting economic, employment, and scientific markets. To ad- of careers are largely determined by fluctuations and not by per- shifting economic, employment, and scientific markets. To ad- sistence. dress these changes, universities and other research institutes sistence. dress these changes, universities and other research institutes Fig. 4 shows the MC results for π 1. For c 1 we observe a have shifted away from tenure at all levels of academia in the last ≥ Fig. 4 shows the MC results for π 1. For c 1 we observe a have shifted away from tenure at all levels of academia in the last drastic shift in the career longevity¼ distribution P L , which thirty years towards meeting staff needs with short-term and non- ≥ drastic shift in the career longevity¼ distribution P L , which thirty years towards meeting staff needs with short-term and non- becomes heavily right-skewed with most careers terminatingð Þ ex- tenure track positions (3). becomes heavily right-skewed with most careers terminatingð Þ ex- tenure track positions (3). tremely early. This observation is consistent with the predictions For knowledge intensive domains, production is characterized For knowledge intensive domains, production is characterized of an analytically solvable Matthew effect model (16) which de- by long-term spillovers bothtremely through early. time This and observation through the is knowl- consistent with the predictions by long-term spillovers both through time and through the knowl- monstrates that many careers have difficulty making forward pro- edge network of associatedof an ideas analytically and agents. solvable A potential Matthew draw- effect model (16) which de- edge network of associated ideas and agents. A potential draw- gress due to the relative disadvantage associated with early career back of professions designedmonstrates around that short-term many careers contracts have is difficulty that making forward pro- back of professions designed around short-term contracts is that inexperience. However, due to the nature of zero-sum competi- there is an implicit expectationgress due of to sustained the relative annual disadvantage production associated with early career there is an implicit expectation of sustained annual production tion, there are a few big winners who survive for the entire that effectively discountsinexperience. the cumulative However, achievements due to of the the nature inof zero-sum competi- “ ” that effectively discounts the cumulative achievements of the in- duration T and who acquire a majority of the opportunities al- dividual. Consequently,tion, there there is a possibility are a few that“big short-term winners” con-who survive for the entire located during the evolution of the system. Quantitatively, the tracts may reduce the incentivesduration forT and a young who scientist acquire to a majorityinvest in of the opportunities al- dividual. Consequently, there is a possibility that short-term con- distribution P N becomes extremely heavy-tailed due to agents human and social capitallocated accumulation. during the Moreover, evolution we of highlight the system. Quantitatively, the tracts may reduce the incentives for a young scientist to invest in ð Þ the importance of an employmentdistribution P relationshipN becomes that extremely is able to heavy-tailed due to agents human and social capital accumulation. Moreover, we highlight with α > 2 corresponding to extreme accelerating career growth. APPLIED ð Þ the importance of an employment relationship that is able to APPLIED Despite the fact that all the agents are endowed initially with the combine positive competitivewith α pressure> 2 corresponding with adequate to extreme safeguards acceleratingMATHEMATICS career growth. same production potential, some agents emerge as superstars to protect against careerDespite hazards the and fact endogenous that all the production agents are un- endowed initially with the combine positive competitive pressure with adequate safeguards MATHEMATICS following stochastic fluctuations at relatively early stages of the certainty an individual issame likely production to encounter potential, in his/her some career. agents emerge as superstars to protect against career hazards and endogenous production un- career, thus reaping the full benefits of cumulative advantage. In an attempt to renderfollowing a more stochastic objective fluctuations review process at relatively early stages of the certainty an individual is likely to encounter in his/her career. for tenure and other lifetimecareer, achievement thus reaping awards, the full quantitative benefits of cumulative advantage. In an attempt to render a more objective review process Discussion measures for scientific publication impact are increasing in use for tenure and other lifetime achievement awards, quantitative An ongoing debate involving academics, university administra- and variety (17–20, 24,Discussion 27, 46, 47). However, many quantifiable measures for scientific publication impact are increasing in use tion, and educational policy makers concerns the definition benchmarks such as theAnh-index ongoing (17) debate do not involving take into academics, account university administra- and variety (17–20, 24, 27, 46, 47). However, many quantifiable of professorship and the case for lifetime tenure, as changes collaboration size or disciplinetion, and specific educational factors. policy Measures makers for concerns the definition benchmarks such as the h-index (17) do not take into account in the economics of university growth have now placed tenure the comparison of scientificof professorship achievement shouldand the at case least foraccount lifetime tenure,ECONOMIC SCIENCE as changes collaboration size or discipline specific factors. Measures for in the economics of university growth have now placed tenure the comparison of scientific achievement should at least account ECONOMIC SCIENCE

Petersen et al. PNAS ∣ April 3, 2012 ∣ vol. 109 ∣ no. 14 ∣ 5217 Petersen et al. PNAS ∣ April 3, 2012 ∣ vol. 109 ∣ no. 14 ∣ 5217 Vol 446 | 5 April 2007 | doi:10.1038/nature05670 LETTERS

Quantifying social group evolution Gergely Palla1, Albert-La´szlo´ Baraba´si2 &Institutional Tama´s Vicsek1,3 trends in Science

The rich set of interactions between individuals in society1–7 dominated by single links, whereas the co-authorship data have many results in complex communityemergence structure, capturing highlyof small-world con- dense, highly connected time-dependent neighbourhoods. Furthermore, the links collaboration in networks with nected circles of friends,• families or professional cliques in a social the phone network correspond to instant communication events, cap- network3,7–10. Thanks to frequent changes in the activity and com- turing a relationship as it happens. In contrast, the co-authorship data munication patterns of individuals, the associated social and com- NATURE PHYSICS DOI:10.1038/NPHYS2180the increasing role of team-work| in science munication network is subject to constant evolution7,11–16. Our INSIGHT PROGRESS ARTICLE a Co-authorship b Phone call knowledge of the mechanisms governing the underlying community dynamics is limited, but is essential for a deeper understanding G. Palla, A.-L. Barabasi, T. Vicsek. Quantifying social group of the development and self-optimization200+ of society as years a whole17–22. Pi(0) Pi(1) 0, that is, for a network without disconnected or = = 97 We have developed an algorithm based on clique percolation23,24 singly connected nodes . evolution. Nature 446, 664-667 (2007) that allows us to investigate the time dependence of overlapping (3) In the case of a loop-like NON (for dependences in communities on a large scale, and thus uncover basic relationships one direction) of n Erdős–Rényi networks96, all the links are characterizing community evolution. Our focus is on networks unidirectional, and the no-feedback conditionS. Wuchty, is irrelevant. B. F. IfJones, the B. Uzzi. The increasing dominance capturing the collaboration between scientists and the calls be- initial attack on each network is the same,of teams 1 p, qini 1productioni qn1 q and of knowledge. Science 316, 1036-9 (2007) tween mobile phone users. We find that large groups persist for − − = = ki k, using equations (15) and (16) we obtain that P satisfies Chain-likelongerPaul if A.NON they David. are capableThe Historical Star-like of dynamically NON Origins altering of Tree-like‘Open their Science’: NON member- = ∞P. Stephan. How Economics Shapes Science. ship, suggesting that an ability to change the group composition Figure 6 AnThree essay types on patronage, of loopless NONreputation, composed and of common ﬁve coupled kP results in better adaptability. The behaviour of small groups dis- P p(1 e− ∞ )(qP q 1) (19)(Harvard Univ. Press, 2012) networks.agency| All have contracting the same percolation in the scientific threshold revolution and the same. giant ∞ = − ∞ − + plays the opposite tendency—the condition for stability is that c d 0.6 component.theirCapitalism composition The dark and node remainsSociety represents unchanged.3(2): the Article origin We network5 also(2008). show on which that know- failures 14 12Note that if q Zip-code1 equation0.5 (19) has onlyZip-code a trivial solution ledge of the time commitment of members to a given community 〉 = Age Age initially occur. 10 0.4 P 0, whereas for q 0 its yields the known giant component rand

n ∞ /

can be used for estimating the community’s lifetime. These find- 〈 8 = = of a single network, equation〉 0.3 (12), as expected. We present /

organizational〉 shifts in the business structure of ings offer insight into the fundamental differences between the 6 real

n 0.2

NON, (2) a tree-like random regular fully dependent NON, (3) a 〈 • real numerical solutions of equation (19) for two values of q in 4

dynamics of small groups and large institutions. n loop-like Erdős–Rényi partially dependent NON and (4) a random 〈 0.1 The data sets we consider are (1) the monthlyresearch list of articles in theuniversitiesFig.2 7b. Interestingly, whereas for q 1 and tree-like structures 0 0 = regularCornell network University of partially Library e-print dependent condensed Erdő matters–Rényi (cond-mat) networks. equations02040 (17) 60 and80 (18) 100 depend 120 0 on 20n, for 40 loop-like 60 80 NON 100 120 structures All casesarchive represent spanning different 142 months, generalizations with over 30,000 of percolation authors25, and theory (2) equation (19)s is independent of n. s for athe single record network. of phone calls In all between examples the customers except of (3) a mobile we apply phone the efGrowth(4) For NONsContraction where each ER network is dependent on exactly no-feedbackcompany condition. spanning 52 weeks (accumulated overshifts two-week-long away per- fromm other Erdtenureős–Rényi networks towards (the case ofshorter-term a random regular (1)iods), We and solve containing explicitly the96 communicationthe case• of a patterns tree-like of over NON 4 million (Fig. 6) ttnetwork + 1 of Erdőtts–Rényi + 1 networks), we assume that the initial attack Merging Splitting formedusers. by Bothn Erd typesős–Rényi of collaboration networks events92–94 (awith new articlethe same or a phone average on each network is 1 p, andtt each partially ∪ t + 1 dependentt + 1 pair has call) document the presence of social interactioncontracts between the + bottle neck− in the number of tenure- degrees k, p1 p, pi 1 for i 1 and qij 1 (fully interdependent). the same q in both directions. The n equations of equation (15) involved= individuals= (nodes),= and can= be represented as (time- tt + 1 t t + 1 Fromdependent) equations links. (15) Theand extraction (16) we obtain of the changingan exact linkexpression weights fromfor the areBirth exactly the sameDeath owing to symmetries, and hence P can be order parameter, the size of the mutual gianttrack component positions for all p, k available ∞ the primary data is described in Supplementary Information. In ttobtained + 1 analytically,tt + 1 and nFig.values, 1a, b we show the local structure at a given time step in the Figure 1 | Structure andp schematic dynamics of the two networks two networks in the vicinity of a randomly chosen individual kP 2 m n considered.P a, The co-authorship(1 e− ∞ network.) 1 q The figure(1 showsq) the4qP local (20) (marked by a redP frame).p 1 Theexp( communitieskP ) (social groups repre-(17) ∞ = 2m − [ − + − + ∞] ∞ = [ − − ∞redefining] thecommunity role structure atof a given teaching time step in the vicinity of-vs- a randomly research selected faculty sented by more densely interconnected• parts within a network of node. b, As a but for the phone-call network. c, The filled black symbols Equationsocial (17) links) generalizes are colour known coded, soresults that forblackn nodes/edges1,2. For n do not1, we from which we obtain = = correspond to the average size of the largest subset of members with the same obtainbelong the known to any community, result pc 1 and/k, those equation that simultaneously (11), of an Erd belongős–Rényi to zip-code, Ænrealæ, in the phone-call communities divided by the same quantity two or more communities= are shown in red. found in random sets, n , as a function of the community size, s. Similarly, network and P (pc) 0, which corresponds to a continuous Æ randæ 1 The two networks∞ = have rather different localshifts structure: thein collab- thethe competitive open symbols show the average sizepaspects of the largest subset of of community science,(21) second-order phase transition. Substituting• n 2 in equation (17) c m oration network of scientists emerges as a one-mode= projection of the members with an age falling in a three-year= k time(1 window,q) divided by the same yields the exact results of ref. 73. − bipartite graph between authors and papers,universities, so it is quite dense and quantity and in random scientists: sets. The error bars in bothreputation cases correspond to Ænreal æ/tournaments Solutionsthe overlap of between equation communities (17) are shown is very significant. in Fig. 7a In for contrast, several in values the (ÆnAgain,randæ 1 srand as) in and caseÆnreal (3),æ/(Æn itrand isæ 2 surprisingsrand), where thatsrand bothis the the standard critical threshold deviation in the case of the random sets. d, The Æn æ/s as a function of s, for of n.phone-call The special network case n the1 communities is the known are Erd lesső interconnecteds–Rényi second-order and are and the giant component are independentreal of the number of percolation law, equation= (12), for a singlein network. omnipresent In contrast, both the zip-code (online) (filled black symbols)and competition theage(open symbols). e, Possible arenas often separated by one or more inter-community nodes/edges. Indeed, eventsnetworks in communityn, in evolution. contrastf, to The tree-like identification NON of evolving (equations communities. (17) and (18)), > for anywhereasn the1, the phone solution record ofcaptures (17) yieldsthe communication a first-order between percolation two Thebut links dependat t (blue) and on the the links couplingat t 1 1 (yellow)q areand merged on into both a joint degrees graph k and people, the publication record assigns to all individuals that contribute transition, that is, a discontinuity of P at pc. (green).m. Numerical Any CPM community solutions at t or oft 1 equation1 is part of a (20) CPM community are shown in the in Fig. 7c, Ourto a results paper a show fully connected (Fig. 7a) clique. that the As∞ a NON result, becomes the phone more data are vul- joined graph, so these can be used to match the two sets of communities. and the critical thresholds pc in Fig. 7c coincide with the nerable1Statistical with and increasing Biological Physicsn Researchor decreasing Group of the HAS,k Pa(´pzmac íncreasesny P. stny. 1A, H-1117 when Budapest, Hungary. 2Center for Complex Network Research and Departments of Physics and 3 theory, equation (21). n increasesComputer Science, or k Universitydecreases). of Notre Furthermore, Dame, Indiana 46566, for USA. a fixedDepartmentn, of when Biological Physics, Eo¨tvo¨s University, Pa´zmańy P. stny. 1A, H-1117 Budapest, Hungary. k is664 smaller than a critical number kmin(n), pc 1, meaning Remark on scale-free networks ≥© 2007 Nature Publishing Group that for k < kmin(n) the NON will collapse even if a single 96 The above examples regarding Erdős–Rényi and random regular node fails . networks have been selected because they can be explicitly (2) In the case of a tree-like network of interdependent random 97 solved analytically. In principle, the generating function formalism regular networks , where the degree k of each node in each network presented here can be applied to randomly connected networks is assumed to be the same, we obtain an exact expression for the with any degree distribution. The analysis of the scale-free networks order parameter, the size of the mutual giant component for all λ with a power-law degree distribution P(k) k− is extremely p, k and n values, important, because many real networks can∼ be approximated n k 1 k by a power-law degree distribution, such as the Internet, the 1 −k n 1 n 1 − P airline network and social-contact networks, such as networks P p 1 p n P n 1 ∞ 1 1 (18) 2,10,51 ∞   ∞   of scientific collaboration . Analysis of fully interdependent = −  − p − + 73     scale-free networks shows that, for interdependent scale-free   networks, pc > 0 even in the case λ 3 for which in a single Numerical solutions of equation (18) are in excellent agreement network p 0. In general, for fully≤ interdependent networks, c = with simulations. Comparing with the results of the tree-like the broader the degree distribution the greater pc for networks Erdős–Rényi NON, we find that the robustness of n interdependent with the same average degree73. This means that networks with a random regular networks of degree k is significantly higher than broad degree distribution become less robust than networks with that of the n interdependent Erdős–Rényi networks of average a narrow degree distribution. This trend is the opposite of the degree k. Moreover, whereas for an Erdős–Rényi NON there exists trend found in non-interacting isolated networks. The explanation a critical minimum average degree k k that increases with n of this phenomenon is related to the fact that in randomly = min (below which the system collapses), there is no such analogous kmin interdependent networks the hubs in one network may depend on for the random regular NON system. For any k > 2, the random poorly connected nodes in another. Thus the removal of a randomly regular NON is stable, that is, pc < 1. In general, this is correct selected node in one network may cause a failure of a hub in for any network with any degree distribution, Pi(k), such that a second network, which in turn renders many singly connected

NATURE PHYSICS VOL 8 JANUARY 2012 www.nature.com/naturephysics 45 | | | NEWSFOCUS this kindNEWS ofNEWS scienceFOCUS is actuallyFOCUS done, if the award had been made collec- cists at LBNL, the tively to all members of the two groups,” Rees told Reuters. SCP began as an Within hours of the announcement, Schmidt and Riess decided to effort to fi nd nearby invite the remaining 17 members of the High-z team to Stockholm for supernovae using an the Nobel ceremony. Each laureate would be allowed 14 tickets to the Who Invented theautomated search various events organized by theNEWS SwedishFOCUS Academy, and between the technique. The tech- two of them, Schmidt and Riess had enough tickets to accommodate Higgs Boson? nique involved tak- everybody and their spouses. The spare tickets they gave to Perlmutter, Festivities. Receptions for Nobelists and hundreds ing telescopic images who had a bigger challenge with the 30 collaborators that he wanted of other guests began days before the ceremony. ofShiny. the sameKing Abdulaziz swaths of University’s steps to gain to invite. By December, all arrangements had been made to bring both skyvisibility at different are controversial. times teams to the world’s grandest scientifi c celebration, with the three lau- and using an algorithm to contrast those images to spot supernovae that reates spending roughly $100,000 from the $1.5 million prize to pay for might have exploded in the time between two shots.tutions. In 1988, All have the group changed their affi liation on their guests’ airfares, hotel rooms,POLICY tuxedo rentals, FORUMand other expenses. proposed applying the technique to fi nd distant supernovae.ISI’s highly Ascited outsid- list—as required by KAU’s After years of a deep and sometimes hostile rivalry, the two groups ers to astronomy, Pennypacker and Muller facedcontract—and a constant challenge some have added KAU as an would have a chance to revel in their shared glory, sip champagne side in getting funded. For this, they would later blameaffi a prominent liation on researchmember papers. Other require- SCIENCE POLICY by side, and possibly reconcile their warring narratives of the discovery of the yet-to-be-formed High-z team: Kirshner, mentswho by in virtuethe contract of his include devoting “the whole of your time, attention, skill and abili- in a scientifi c colloquium at the end of the celebrations. supernova expertise was on proposal review committeesNational incentiveappointed policies by relate to increases ties to the performance of your duties” and the Department of Energy and the National Sciencein research Foundation. article submissions and Changing Incentives to Publish publicationsdoing “work in equivalent Science. to a total of 4 months Monday, 5 December By 1991, Pennypacker’s interests had turned perto science contract education, period.” Chiara Franzoni,1 Giuseppe Scellato, 2 ,3 Paula Stephanand Muller 4 ,5 ,6* had shifted to studying weather patterns.Neil The Robertson, two handed a professor emeritus the reins of the SCP to Perlmutter—a hawk-nosed,of mathematics tenacious, young at Ohio State University in physicist who had been Muller’s graduate student. Perlmutter’s impres- any national governments have Columbus who has signed on, says he has December is bleak in Stockholm. On most days, the sun sets at 2:00 sive organizational4 skills helped seal his positionno asconcerns the undisputed about Trendthe offer. in number “It’s just of submit-capi- implemented policies providing Institutional incentives p.m., enveloping the city in a darknessCITATION that seems IMPACT merciful at the end leader of the project, even though the group includedtalism,” a senior, he says. and ted “They at papers have to Science the capital. Thirty Five living theorists haveincentives claims to for having researchers dreamed to pub- up the most3.5 famousIndividual–career Kingdom. incentives Others question whether the

countries on September 14, 2012 by type of incentive of what has usually been a gray, M overcast morning. The joke among the time, moreIndividual–cash distinguished, bonus physicist named Gersonand they Goldhaber. want to build something out of it.” advance was a big enough step beyond pre- lish, especially in highly ranked international relative to number submitted guests attendingsubatomic the Nobel particle festivities in physics. is that Butthe Swedeswhat did invented they really the do?Perlmutter3 No systematized policy changevious work the search to merit technique. science’s biggest AnotherHe demonstrated prize. KAU affiliate, that astronomer Gerry journals. Although still the top publishing by country in 2000 [see (13 ) Nobel Prize to bring cheer to StockholmSaudi in itsUniversities darkest month and one couldOffer more or Cashless guaranteeIn any case, fi nding says Chris supernovae Quigg,Gilmore aby theorist taking of the a University refer- of Cambridge in nation, the United States has seen its share Science 2.5 and SOM for details]. boost the NOWlocal economy THAT THE with HIGGS an infl BOSON—ORux of tourists. standard model, the Higgsence boson image itself of is ain patch a at of Fermi the skyNational just after Accelerator a newthe moonLaboratory United and Kingdom, subtract- notes that “universities of publications decline from 34.2% in 1995 The twosomething teams began much likearriving it—is inin thethe bag,city theon 5 December.way the byproduct All of of moreing importantit 2from another under- image(Fermilab) of the in same Batavia, sky takenIllinois, buyright historians people’s before reputations the next all the time. In prin- to 27.6%In inExchange 2007 as the number of forarticles Academic Prestige recruits and to decide ten- the High-zquestion members on manyhad rooms people’s reserved minds is at who the magnifilying physics. cent Grand Instead ofnew the boiler moon. on Through a steam theand early prize 1990s, committees Perlmutter must be expanded carefulciple, this to givethe is no group different by from Harvard hiring gets the Nobel Prize publishedfor the discovery. by U.S. scientistslocomotive, and engineers it’s more like the1.5 whistle: Its toot credit precisely where and for what it is due: ure and promotions. In Hotel, where laureates stay. TheTwo Grand Saudi was institutions already full are by aggressively the time acquiringrecruiting the collaborators affi liations inof Europeoverseas and scientists Australia. aWhat prominent had begun researcher.” as a If you go by the pophas history,plateauedwith an the eye and answer to that gaining ofproves other visibility thatcountries the in boiler research has is there journals and working, “If these people receive their rewards,Officials either at KAUGermany, did not reforms respond were to Submissions to the SCP teamis obvious. made Inreservations, 1964, Peter soHiggs, its members a mild- hadbut itto doesn’t fi nd rooms turn the wheels.team of1 physicists grewin heavento include or before, several it would astronomers. be nice if itBut was the group on December 12, 2011 grown (1 , 2). Hicks ( 3) argues that the two Science’s requestmade forwww.sciencemag.org anthat interview. allow univer- But elsewhere.mannered “We were theorist a bit latefrom off the the University gate,” says of AndrewAs for Fruchter, whether thea workstill of had Higgs a tough and col- time persuadingfor something review they committeesactually did and of telescope not for facili- eventsAt arefirst not glance, unrelated: Robert The Kirshner decline tookin the the 4 years largely through initiatives specifi cally Surender Jain, a retiredsities tomathematics link salaries pro- to member ofEdinburgh Perlmutter’s in the group.United Kingdom, dreamed leagues merits a Nobel tiesPrize, to0.5 opinionsgrant them vary. observing what people time. say they did.” up the particle to explainrelative thee-mail origins performance message of mass. for of a“Certainly, scam.the United An yes,astronomer States I think it istargeted at least astoward pro- attaching KSU’s name to fessor from Ohio Universityresearch in performance Athens who In the race that led up to therelates discoveryat Kingto increased Abdulaziz of the international accelerating University competition (KAU)universe, in Jed- Whileresearch0 the publications, SCP was led regardless by physicists of whether interested is an in adviser astronomy to KAU as(table aand S1, has SOM). helped recruit He completed physicists’ standard model of found as other things that have been2000 given2001 the 2002The 2003problem 2004 2005 2006 2007 2008 2009 however, Perlmutter’s group engenderedhad dah,been Saudi the by fi newlyArabia,rst to start.adopted was Founded offering incentives himin the thata con- tool theto understandwork involved the any universe, meaningful the High-z collabora- collaborationseveral of grew the adjuncts,out of aSome provided countries a list of have61 early 1980sfundamental by Carl Pennypackerparticles and forces.and Richard Experi- Muller,Nobel bothin the physi-past,” says teamFrank of Close, astronomers a theo- whoWhat realizedHiggs andYear that company Type 1a did, supernova albeit unwit- explosions menters working withhave thetract crowded world’s for an out largestadjunct some professorshipworkrist at by the U.S. University authors.that would of Oxford tion in with the UnitedKSU researchers. tingly, was to dynamite a hugeacademics boulder that who haveintroduced signed contracts a system simi- of atom smasher, the LargeWepay Hadron investigate $72,000 Collider a howyear. changes Kirshner, in an incentives astrophysi- couldhave helpAcademicsbeen them less stronganswer both inside with a fundamental andregard outside to salary Saudiphysicswas blocking cash larquestion: to progress bonusesthe one the sent tofate individualsto of Kirshner. Thefor eachfi nancial arti- Institutional context: Increasing teamDownloaded from size & changing incentive system (LHC) at the Europeanto publishparticlecist at Harvardphysicsimplemented lab- University, at the would country be expected level the and cosmos.Arabia promotion. warn These that Funding astronomers—including suchOnline practices for research could detract oftenon the Mario standardclearrangements publishedHamuy, model Nicholas in thea top contracts international vary, Jain scientifi says: c oratory, CERN, in Switzerland, have now Suntzeff, Mark Phillips, and others—hadof been particle studying physics. nearby Type on September 19, 2011 relateto to supervise the number a research of submissions group at and KAU pub- and didfrom not theemphasize genuine effortspublications that Saudi in interna- Arabia’s journal.For instance, Turkey some introduced adjuncts inwill 2008 receive a national their www.sciencemag.org Science sciencemag.org seen that particle ( licationsspend, 13 andJuly, a weekthe p. 141).acceptance or two a year rates on to KAU’s the jour- cam-1ational supernovaeuniversities journals. for are Departments years making before to transform oftenthey beganreceived them-The the standard searchagencycompensation model for that distant iscollects not Type as publication salary but asdata part and, of fora So Higgs gets the glory. Podcast interview a quantum fi eld theory, nal Sciencepus, but for that 27 requirement OECD (Organization was fl exible, for the 1a funds supernovae.selves based into world-class onBecause enrollment the withresearch universe writer numbers Adrian centers. is expanding, and For eachresearch far-off article, grant supernovae pays provided a cash by bonus KAU. equivalent to Only that’s not exactly what happened. In Cho (http://scim.ag/ so it focuses on quan- Economicperson Cooperationmaking the offer and wrote Development) in the e-mail. recede numberinstance, from of personnel.theEarth Saudi at governmentsuch great velocitieshas spent bil- that their~7.5%Jain light ofacknowledges thereaches average us that faculty a primary salary goal (7 ,of 8 ). fact, theorists say, Higgs madeWhat a fairly Kirshner narrow would be required to do, lions of dollars to buildpod_6100). the new King Abdul-tum wavesthe orprogram—funded “fi elds” by Saudi Arabia’s Min- and esoteric advancecountries in mathematical andNEWS NEWS3 OECD-monitoredphys-FOCUSFOCUS countries stretchedIncentives in wavelengths to publish in toward international the red jour-that end describe ofThe the Chinesetheelectromagnetic prob- Academy of Sciences adopted a however, was add King Abdulaziz Univer- lah University of Science and Technology in istry of Higher Education—is to “improve ics. Several other physicists(China, madeRussia, the and same Singapore) for the period spectrum—anals beganFOUND to “redshift”be moreability widespread represented of fi nding in the byvarious 1980s. the letterparticles bonus z. That’shere policy and why in 2001.these Rewards vary by insti- sity as a second affi liation to his name on the Thuwal, which boasts state-of-the-art labs the visibility and ranking of King Abdulaziz advance at the same 2000–09.time. Their We intellectual further differentiate by type of objectsWhoIn some are Inventedcountries, known as incentives there.high-redshift the It contains apply onlyor fi eldshigh-z to sci-for supernovae.the dozentute but types represent Unlike of Perl-a large amount of cash com- leap was essential to the developmentInstitute for of Scientifi the c Information’s (ISI’s) and dozens of prominentmatter particles—includingresearchers as full- electrons,University.” the up But he says KAU also hopes the incentive. Our analysis shows that the intro- mutter’sence and group, engineering; the High-z in other team countries, was a fl atthey organization. pared with Even the standardthough salary of the research- Downloaded from list of highly cited researchers. Higgstime faculty Boson? members (Science, 16 October foreign academics will help it kick-startEmergence indig- of “big science” standard model, perhapsduction the mostof incentives elaborate by a country is associ- Schmidtapply to wasa wider technically rangequarks of disciplines.the and leader, down Thequarks the UK team that ers.makewas Bonuses aup collaboration pro- are particularly high for publica- and precise theory in all of science.“I thought But their it was a joke,” says Kirshner, 2009, p. 354). tons and neutrons, and the heavierenous analogsresearch programs. “We’re not just giv- www.sciencemag.org ated with an increase in submissions by the amongtook the equals, lead with adoptiondifferent membersof the Research getting primarytions0 in authorshipjournals such on as Science and Nature 19581962 who forwarded the e-mail to his department But the initiatives at KSU and KAU are 10ing away money,” he says. Most recruits will papers didn’t even mention the most impor- papers that they individuallyof those ledparticles about that different emerge aspectsin9 high-energy of the work. Nature/PNAS/Science 1963 1967 tant problem their workcountry; helpedchair, to the notingsolve. relation Other in jest is that particularly the money strongwas a lot Assessmentaimed at getting Exercise speedierparticle (RAE) collisions.results. in 1986, “They which are (be ). expected The Korean to visit government for a total of inaugurated 4 weeks in a a 19681972 In 1993, the year before the team began taking those high-redshift scientists did that later—butbetweenmore their cash attractive contribu- bonuses than and the 2%submissions. annual raise We pro- allocatessimply nationalbuying names,” funds Theseto says departments Mohammedparticles oninteract theAl- throughsimilaryear to “give policythree crash in 2006 courses”; whereby they 3 will million also wonbe 1973 1977 Founders. 1 19781982 CREDITS (TOP TO BOTTOM): NICHOLAS SUNTZEFF; CARLTON PENNYPACKER; ROBERT KIRSHNER; COURTESY OF A. DIERCKS; COURTESY OF A. FILIPPENKO; MATT CASHORE/UNIVERSITY OF NOTRE DAME; COURTESY R. GILLILAND; REIDAR GILLILAND; R. COURTESY DAME; NOTRE OF CASHORE/UNIVERSITY MATT FILIPPENKO; A. OF COURTESY DIERCKS; A. OF COURTESY KIRSHNER; ROBERT PENNYPACKER; CARLTON SUNTZEFF; NICHOLAS BOTTOM): TO (TOP CREDITS HEYER/ESO HERMANN HANS KIRSHNER; R. COURTESY UNIVERSITY; ACEVEDO/RUTGERS MIGUEL HAHN; Pennypackertion (which won(left) Nobelandfi Mullernd laurels some (third in indication )1979) ceded stillSCP that to Perlmutter publications (second relate). observationsbasis of past fromperformance theforces: Cerro and the Tololo peerelectromagnetic review. Inter-American A force(roughly that Observatory binds U.S. $2800) in is paid to the fi rst and fessors typically get. Then he discovered that Qunaibet, a professor of agricultural eco-10expected to supervise dissertations and help 19831987 Q: how to “fairly” doesn’t explain the origins of all mass. the atom, the strong nuclear force that binds

to career-based incentives. number of factors are included in the rank- a) corresponding authors on papers in key jour- a highlyNEWS citedFOCUS colleague at another U.S. insti- nomics at KSU, who recently criticized the KAU’s full-time faculty members develop 19881992

Even the famous particle, the Higgs boson, quarks into protons and! neutrons, and theScience Nature Cell 10 19931997 tution had accepted KAU’s offer, adding KAU ings,programs but publications in an article constitute he wrote for the the core leading for nalsresearch2 such proposals.as ,Even the, and “shadows” ( ).of doesn’t quite live up Incentivesto its legendary status. Name recognition. Peter Higgsscience was one and of six engineering theo- weak (5 nuclear, 6). force, which10 produces a kind 19982002 distribute credit in a as a second affi liation on ISIhighlycited.com. Saudi newspaper, Al Hayat. Teddi Fishman, such eminent scholars would inspire local stu- 20032007 Downloaded from High-z Often portrayed as the engine that drives the rists to have the same idea. of radioactivity. (The standard model does BALIBOUSE/REUTERS/LANDOV DENIS ATLAS; BOTTOM): TO (TOP CREDIT Five living theorists have claims to having dreamed up the most famous Kingdom. Others question whether the Shiny. King Abdulaziz University’s steps to gain IncentivesKirshner’s for faculty colleague to publish is not have alone. a long Sci- directorThe UK of reform the Center provided for Academic an example on September 14, 2012 Integ- for Datadents and and Modelsfaculty members, he says. 20082012 advance was a big enough step beyond pre- visibility are controversial. subatomic particle in physics. But what did they really do? vious work to merit science’s biggest prize. Supernova P (A Science historyence in has the learned United of moreStates than and 60 Canada. top-ranked governmentsrity at ClemsonIn anyworldwide. case, University says Chris Quigg, Australia a theoristin South and Carolina, New We3 Thestudied recruits the journalScience spoke to because say they of system dominated by 1286 14NOW SEPTEMBER THAT THE HIGGS 2012 BOSON—OR VOLstandard 337 model, the SCIENCEHiggs boson itself is in awww.sciencemag.org at Fermi National Accelerator Laboratory 10 Search Team Promotionresearchers andsomething tenure,from much different like as it—is well in the bag,scientific as the compen-way the byproduct disci- of moreZealand importantsays under- the drew programs(Fermilab) on in Batavia, the deliberately Illinois,RAE historians to createput in tutions.“a place false All haveitshave highchanged a genuine impact their interestaffi factor liation in andon promoting international research and question on many people’s minds is who lyingP uphysics.blish Insteaded b yof AtheA boilerAS on a steam and prize committees must be careful to give ISI’s highly cited list—as required by KAU’s Members sation,plines—all dependgets toon the aNobel ISI’sconsiderable Prize forhighly the discovery. cited extentlocomotive, list—who on it’s amore like policy the whistle:impression Its reforms toot credit thatprecisely for these where funding and universitiesfor what itacademic is due: are institu-produc- interdisciplinaryat KAU, even though scope. none Moreover, of them knew the annual how ▲ If you go by the pop history, the answer proves that the boiler is there and working, “If these people receive their rewards, either contract—and some have added KAU as an teams? is obvious. In 1964, Peter Higgs, a mild- but it4 doesn’t turn the wheels. in heaven or before, it would be nice if it was facultyhave member’s recently signedpublication a part-time record employment ( ). An tionsing wherebygreat research.” better-performing institutionswww.sciencemag.org numbertheir4 individual of published research articles plans wouldhas remained match P. Challis A. Diercks manneredA. theoristFilippenko from the University P.of GarnavichAs for whether the work ofR. Higgs Gilliland and col- for something theyC. actually Hogan did and not for S. affiJha liationCDF, 10 on researchR. Kirshner papers. OtherB. Leibundgut require- arrangementEdinburgh with in the the United university Kingdom, dreamed thatleagues is merits struc- a Nobel Prize, opinionsAcademics vary. what people say who they did.” have accepted KAU’s up with the interests and abilities of KAU’s active labor marketup the particle exists to explain thefor origins highly of mass. “Certainly,produc- yes, I think itreceive is at least as pro-more funding than lower-performingments in thefairly contract constant include devoting at ~800. “the During the 10-year tured alongHe completedthe lines physicists’ of standard what model ofKirshner found as other thingswas that have offerbeen given therepresent The problem a wide variety of wholefaculty of yourfaculty time, attention, members skill and and students.abili- Ray Carlberg, tive faculty,“50-way whofundamental often particles increase and forces. tie Experi- their Nobel forsalaries in the past,” the says Frankones Close, Nobel and, a theo- thus,What Higgs have and companyPrize” more did, albeit resources unwit- to com- study period, fi rst authors from 144 differ- Q: how to offered. Meanwhile,menters working with a the bigger, world’s largest more rist at the promi- University of Oxfordfrom in the United elite tingly, institutions was to dynamite a huge in boulder the that United tiesStates, to the performancean astronomer of your at duties” the University and of Toronto in by receiving offersatom smasher, from the Large alternative Hadron Collider institu- pete in the job marketwas for blocking scientists. progress Norway, ent5 countries submitted 110,870 original SCIENCE Downloaded from 27 www.sciencemag.org(LHC) at the European particle physics lab- VOL 336Online 6 APRILon the standard2012 model doing “work10 equivalent0 to a total of 4 months 1 2 3 tions.nent In many Saudi oratory,other institution—King CERN, countries, in Switzerland, have incentives now Saud Univer- for Belgium,Canada, Denmark, Europe, Asia,andof particle Italy and physics. started Australia. similar All researchCanada who articles; accepted 7.3% the of offer,these sayssubmissions he had seen that particle (Science, 13 July, p. 141). Published by AAAS sciencemag.org The standard model is per contract period.”10 10 10 10 sity in Riyadh—hasSo Higgs gets the glory. climbed several hundred are men. SomePodcast interview are a emeritusquantum fi eld theory, professors who to Google the university after he received the faculty to publish in international journals policies duringwith writer the Adrian past decade for allocatingNeil Robertson,were accepted a professor for emeritus publication, with first “fairly” Only that’s not exactly what happened. In Cho (http://scim.ag/ so it focuses on quan- places in internationalfact, theorists say, Higgs made rankings a fairly narrow in the past a sharehave recentlyof pod_6100).the budget retiredtum [tablewavesfrom or “fi theirelds”S1, supportinghomeof insti-mathematics authorse-mail. at Ohio Hefrom State admits 53 University different Coauthorsthat he in countrieswas initially (11 per ,con- 12). paper,COMMONS SHAKER/WIKIMEDIA AMMAR CREDIT: a and esoteric advance in mathematical phys- that describe the prob- Columbus who has signed on, says he has 1Department of Industrialics. Several otherEngineering physicists made andthe same Management, ability of fi nding various particles here and advance at the same time. Their intellectual online materialthere. It contains (SOM)]. fi elds for the dozen types of no concerns aboutWe the analyzed offer. “It’s funding just capi- and reward policies Politecnico di Milano,leap was 20133 essential toMilan, the development Italy. of2 Departmentthe of matter particles—including electrons, the up distribute 1344 CITATIONstandard model, IMPACT perhaps the most elaborate 9 DECEMBEROther 2011 quarks countries VOL and down 334 quarks focus that SCIENCE make up pro- on incentives www.sciencemag.orgtalism,” he for says. 30 “They countries, have the which capital collectively repre- Production Systemsand and precise Business theory in all Economics,of science. But their Politecnico di tons and neutrons, and the heavier analogs and they want to build something5-year out of it.” growth rate = 0.064(2) 0.013 annual growth Torino, 10129 Turin,papers Italy. didn’t 3Bureau even mention of Researchthe most impor- on Innovation, directed atof thoseindividuals particlesPu bthatlis emergehed binrather yhigh-energy AAAS than institu- sent 95% of all articles submitted and 99% tant problem their work helped to solve. Other particle collisions. Another KAU affiliate, astronomer Gerry Complexity and Knowledge,scientists did that Collegio later—but theirCarlo contribu- Alberto, 10024 tions. GermanyThese particles and interact Spain through made three reforms in of all articles published in Science during the Saudition (which won NobelUniversities laurels in 1979) still Offerforces: Cash the electromagnetic force that binds Gilmore of the University of Cambridge in credit in a Moncalieri, Italy. 4Andrewdoesn’t explain Young the origins School of all mass. of Policy Studies, the atom, the strong nuclear force that binds 1 rate of the mean Even the famous particle, the Higgs boson, the mechanismsquarks into protonsthat and regulate neutrons, and the access tothe uni-United Kingdom,period (seenotes chartthat “universities and table). Eleven of the 30 Georgia State University, Atlanta, GA 30302, USA. 5Depart- 10 doesn’t quite live up to its legendary status. Name recognition. Peter Higgs was one of six theo- weak nuclear force, which produces a kind buy people’s reputations all the time. In prin- ment of EconomicsIn OftenCognettiExchange portrayed de as theMartiis, engine that University drives the ristsfor of to haveTorino, the sameAcademic idea. versity careers,of radioactivity. promotion, Prestige (The standard model and does BALIBOUSE/REUTERS/LANDOV DENIS ATLAS; BOTTOM): TO (TOP CREDIT salary, link- countries have introduced reforms and poli- 6 ciple, this is no different from Harvard hiring 10124 Turin, Italy.1286 National Bureau of Economic14 SEPTEMBER Research, 2012 VOL 337 SCIENCE www.sciencemag.org Two Saudi institutions are aggressively acquiringing themthe affi more liations tightly of overseas to international scientists a publi- prominent researcher.”cies related to incentives to publish in interna- systemcollaboration size a Cambridge, MA 02138, USA. Published by AAAS ! " with an eye to gaining visibility in research journalscations. In Spain, a national agency wasOfficials put tional at KAU journals did not in respond the past to 10 years. Incentives *Author for correspondence. E-mail: [email protected] in place to assess the performance of Sciencyounge’s requestare subdivided for an interview. into three But categories: on December 12, 2011 policies a • At firstPOLICY glance, Robert Kirshner took the 4 years largely through initiatives specifi cally Surender Jain, a retired mathematics pro- ! " dominated FORUM is consistent with the e-mail message for a scam. An astronomer targeted toward attaching KSU’s name to fessor from Ohio University in Athens who ▲ at King Abdulaziz University (KAU) in Jed- research publications, regardless of whether is an adviser to KAU and has helped recruit !a 702 SCIENCE POLICY 5 AUGUST 2011 VOL 333 SCIENCE www.sciencemag.org dah, Saudi Arabia, was offering him a con- the work involved any meaningful collabora- several of the adjuncts, provided a list of 61 Published by AAAS National incentive0 policies relate to increases Q95 ★ by teams?growth in the grad/ tract for an adjunct professorship that would tion with KSU researchers. academics who Coauthors per paper have10 signed contracts simi- payChanging $72,000 a year. Kirshner, Incentivesan astrophysi- Academics to both Publish inside and outside Saudi lar to thein oneresearch sent toarticle Kirshner. submissions The fi nancial and cist at Harvard University, would be expected Arabia warn that such practices could detract arrangementspublications in the contractsin Science vary,. Jain says: postdoc populations to supervise a research group at KAU and from the genuine efforts that Saudi Arabia’s For instance, some adjuncts will1958- receive their 1968- www.sciencemag.org 1978- 1988- 1998- 2008- Chiara Franzoni,1 Giuseppe Scellato, 2 ,3 Paula Stephan 4 ,5 ,6* spend a week or two a year on KAU’s cam- universities are making to transform them- compensation not as salary but as part of a pus, but that requirement was fl exible, the selves into world-class research centers. For research grant provided by KAU. 1963- 1973- 1983- 1993- 2003- person making the offer wrote in the e-mail. instance, the Saudi government has spent bil- Jain acknowledges that a primary goal of any national governments have 4 What Kirshner would be required to do, lions of dollars to build the new King Abdul- the program—funded by SaudiTrend Arabia’s in number Min- of submit-time period implemented policies providing Institutional incentives ted papers to Science. Thirty however, was add King Abdulaziz Univer- lah University3.5 of Individual–careerScience and Technology incentives in istry of Higher Education—is to “improve incentives for researchers to pub- countries by type of incentive M Individual–cash bonus sity as a second affi liation to his name on the Thuwal, which boasts state-of-the-art labs the visibility and ranking of King Abdulaziz lish, especially in highly ranked international 3 No policy change relative to number submitted Institute for Scientifi c Information’s (ISI’s) and dozens of prominent researchers as full- University.” But he says KAU also hopes the 13 journals. Although still the top publishing by country in 2000 [see ( Downloaded from ) list of highly cited researchers. time faculty members (Science, 16 October foreign academics will help it kick-start indig-

nation, the United States has seen its share Science 2.5 and SOM for details]. “I thought it was a joke,” says Kirshner, 2009, p. 354). enous research programs. “We’re not just giv- of publications decline from 34.2% in 1995 who forwarded the e-mail to his department But the2 initiatives at KSU and KAU are ing away money,” he says. Most recruits will to 27.6% in 2007 as the number of articles recruits and to decide ten- chair, noting in jest that the money was a lot aimed at getting speedier results. “They are be expected to visit for a total of 4 weeks in a morepublished attractive by than U.S. the scientists 2% annual and raise engineers pro- simply buying1.5 names,” says Mohammed Al- year to “give crash courses”;ure they and will promotions.also be In has plateaued and that of other countries has Germany, reforms were

fessors typically get. Then he discovered that Qunaibet,Submissions to a professor of agricultural eco- expected to supervise dissertations and help 1 2 3 1 a highlygrown cited ( , colleague). Hicks at( another) argues U.S. that insti- the twonomics at KSU, who recently criticized the KAU’s full-time faculty membersmade that develop allow univer- tutionevents had areaccepted not unrelated: KAU’s offer, The adding decline KAU in theprograms 0.5 in an article he wrote for the leading research proposals. Even thesities “shadows” to link of salaries to as relativea second afﬁperformance liation on ISIhighlycited.com. of the United StatesSaudi newspaper, Al Hayat. Teddi Fishman, such eminent scholars wouldresearch inspire local performance stu- relates to increased international competition 0 (table S1, SOM). Kirshner’s colleague is not alone. Sci- director of the2000 Center2001 for 2002 Academic 2003 2004 Integ- 2005dents 2006 and 2007 faculty 2008 members, 2009 he says. enceengendered has learned by of newly more adoptedthan 60 top-rankedincentives thatrity at Clemson University in South Carolina,Year The recruits Science spokeSome to say countries they have researchershave crowded from out different some work scientific by U.S. disci- authors.says the programs deliberately create “a false have a genuine interest in promotingintroduced research a system of plines—allWe investigate on ISI’s highly how changes cited list—who in incentives impression have beenthat these less universitiesstrong with are regard produc- to salaryat KAU, cash even thoughbonuses none to ofindividuals them knew forhow each arti- haveto recentlypublish signedimplemented a part-time at the employment country level ing greatand research.”promotion. Funding for research oftentheir individualcle published research in aplans top internationalwould match scientiﬁ c arrangementrelate to the with number the university of submissions that is struc- and pub- Academicsdid not emphasize who have publications accepted KAU’s in interna- up withjournal. the interests Turkey and introduced abilities of in KAU’s 2008 a national on September 19, 2011 turedlications along andthe thelines acceptance of what Kirshner rates to wasthe jour- offertional represent journals. a wide Departments variety of oftenfaculty received faculty membersagency that and collects students. publication Ray Carlberg, data and, for offered.nal Science Meanwhile, for 27 a OECD bigger, (Organizationmore promi- forfrom fundselite institutions based on inenrollment the United numbersStates, anand astronomer each article, at the Universitypays a cash of bonusToronto equivalent in to nentEconomic Saudi institution—King Cooperation and Saud Development) Univer- Canada, number Europe, of personnel. Asia, and Australia. All Canada ~7.5%who accepted of the theaverage offer, facultysays he salaryhad (7 , 8 ). sitycountries in Riyadh—has and 3 OECD-monitoredclimbed several hundred countries are men.Incentives Some are emeritusto publish professors in international who jour-to GoogleThe the Chinese university Academy after he ofreceived Sciences the adopted a

places(China, in international Russia, and Singapore)rankings in for the the past period have nalsrecently began retired to be morefrom widespreadtheir home insti-in the 1980s.e-mail. bonusHe admits policy that in he 2001. was Rewardsinitially con- vary byCOMMONS SHAKER/WIKIMEDIA AMMAR CREDIT: insti- 2000–09. We further differentiate by type of In some countries, incentives apply only to sci- tute but represent a large amount of cash com- 1344 incentive. Our analysis shows that9 DECEMBER the intro- 2011ence VOLand engineering; 334 SCIENCE in other www.sciencemag.org countries, they pared with the standard salary of the research- duction of incentives by a country is associ- apply to aP uwiderblished range by AAA Sof disciplines. The UK ers. Bonuses are particularly high for publica- www.sciencemag.org ated with an increase in submissions by the took the lead with adoption of the Research tions in journals such as Science and Nature country; the relation is particularly strong Assessment Exercise (RAE) in 1986, which ( 9). The Korean government inaugurated a between cash bonuses and submissions. We allocates national funds to departments on the similar policy in 2006 whereby 3 million won fi nd some indication that publications relate basis of past performance and peer review. A (roughly U.S. $2800) is paid to the fi rst and to career-based incentives. number of factors are included in the rank- corresponding authors on papers in key jour- ings, but publications constitute the core for nals such as Science, Nature, and Cell ( 10). Incentives science and engineering (5, 6). Incentives for faculty to publish have a long The UK reform provided an example for Data and Models Downloaded from history in the United States and Canada. governments worldwide. Australia and New We studied the journal Science because of Promotion and tenure, as well as compen- Zealand drew on the RAE to put in place its high impact factor and international and sation, depend to a considerable extent on a policy reforms for funding academic institu- interdisciplinary scope. Moreover, the annual faculty member’s publication record ( 4). An tions whereby better-performing institutions number of published articles has remained active labor market exists for highly produc- receive more funding than lower-performing fairly constant at ~800. During the 10-year tive faculty, who often increase their salaries ones and, thus, have more resources to com- study period, fi rst authors from 144 differ- by receiving offers from alternative institu- pete in the job market for scientists. Norway, ent countries submitted 110,870 original tions. In many other countries, incentives for Belgium, Denmark, and Italy started similar research articles; 7.3% of these submissions faculty to publish in international journals policies during the past decade for allocating were accepted for publication, with first a share of the budget [table S1, supporting authors from 53 different countries (11 , 12). 1Department of Industrial Engineering and Management, online material (SOM)]. We analyzed funding and reward policies Politecnico di Milano, 20133 Milan, Italy. 2Department of Production Systems and Business Economics, Politecnico di Other countries focus on incentives for 30 countries, which collectively repre- Torino, 10129 Turin, Italy. 3Bureau of Research on Innovation, directed at individuals rather than institu- sent 95% of all articles submitted and 99% Complexity and Knowledge, Collegio Carlo Alberto, 10024 tions. Germany and Spain made reforms in of all articles published in Science during the 4 Moncalieri, Italy. Andrew Young School of Policy Studies, the mechanisms that regulate access to uni- period (see chart and table). Eleven of the 30 Georgia State University, Atlanta, GA 30302, USA. 5Depart- ment of Economics Cognetti de Martiis, University of Torino, versity careers, promotion, and salary, link- countries have introduced reforms and poli- 10124 Turin, Italy. 6National Bureau of Economic Research, ing them more tightly to international publi- cies related to incentives to publish in interna- Cambridge, MA 02138, USA. cations. In Spain, a national agency was put tional journals in the past 10 years. Incentives *Author for correspondence. E-mail: [email protected] in place to assess the performance of young are subdivided into three categories: policies

702 5 AUGUST 2011 VOL 333 SCIENCE www.sciencemag.org Published by AAAS Scientific output inflation what is the relative impact/visibility of a publication today -vs- Y years ago? 104 Nature/PNAS/Science Scientific output × 1.5 increase due to technological factors, annual growth population growth, and # papers rate = 0.004 “output inflation”

103 1970 1980 1990 2000 2010

1 10 annual growth rate = 0.014 × 10 growth of team science per paper Ave # coauthors

100 1970 1980 1990 2000 2010 Scientific output inflation what is the relative impact/visibility of a 3 PLoSPLoS One One: publication today -vs- Y years ago? 10 104 monthly Nature/PNAS/Science growth rate # papers 102 gp = 0.048 × 1.5 0 12 24 36 48 60 72 months after first issue annual growth # papers rate = 0.004

103 1970 1980 1990 2000Open2010 Access Journalsgrowth of team science 1 10 annual growth PLoS One: rate = 0.014~ 6,700 articles in 2010 and ~ 14,000 in 2011 × 10 ⇒ × 2 growth in one year alone! per paper ... who is reading/refereeing all these papers?? Ave # coauthors

100 1970 1980 1990 2000 2010 72 The European Physical Journal B

A B

100 > 100 D 90 90

80 80

70 70

60 60

50 50 League average < K > 40 40 Detrended league average < K 1920 1940 1960 1980 2000 1920 1940 1960 1980 2000 Year Year CD14 14 > D 12 12 10 10 8 8 6 6 4 4

League average < HR > 2 2 Detrended league average < HR 1920 1940 1960 1980 2000 1920 1940 1960 1980 2000 Year Year Fig. 4. Acomparisonoftraditionalanddetrendedleagueaveragesdemonstratestheutilityofthedetrendingmethod.Annual per-player averages for (A) strikeouts (B) detrended strikeouts (for pitchers),Accounting (C) home run, for and (D)Inflation detrended home runs (for batters). The detrended home run average is remarkably constant over the 90-year “modern era” period 1920–2009, however there remains a negative trend in the detrended strikeout average. This residual trend in the strikeout average may result from the decreasing role of starters (resulting in shorter stints) and the increased role in the bullpen relievers, which affects the average number of opportunities obtained for players in a given season. This follows from the definition of the detrended average given by equation (10). A second detrending for average innings pitched per game might remove this residual trend demonstrated in Figure 5.Thesharpnegativefluctuationsin1981and1994–1995correspondtoplayerstrikesresultinginJust as the price season stoppage and a reduced average number of opportunities y(t) for these seasons. ! " of a candy bar has increased by a factor of ~ 20 removed (detrended) by normalizing accomplishments by where P is the average of P (t) over the entire period, " # over the last 100 the average prowess for a given season. years (roughly 0.04 2009 We first calculate the prowess Pi(t)ofanindividual 1 3% inflation rate), P Raising the moundP (’62)(t) . (8) player i as 0.035 ≡ 110 " # the home run t=1900 Pi(t) xi(t)/yi(t), (4) 0.03 " PED hitting ability ≡ era The choice0.025 of normalizing with respect to P ofis players arbi- has where x (t)isanindividual’stotalnumberofsuccessesout x 3 i trary, and we could just as well normalize with respectalso increased to by of his/her total number of opportunities y (t)inagiven 0.02 i P (2000), placing all values in terms of current “2000a significant US year t. To compute the league-wide average prowess, we 0.015 dollars”, as is typically done in economics. factor over the then compute the weighted average for season t over all 0.01 Lowering the mound (’69) same period In Figure 4 we compare the seasonal average of x(t) players Home Run Prowess 0.005 end of dead-ball era, emergenceD of " # to the prowess-weighted“Ruthian” average powerx hitters(t) ,forstrikeoutsper 0 " # x (t) player and1900 home1920 runs1940 per player.1960 We1980 define2000 x(t) as P (t) i i = w (t)P (t), (5) Year, t " # " #≡ y (t) i i !i i i 1 " x(t) = xi(t) " # N(t) ! i where " y (t) yi(t) w (t)= i . (6) = P (t) i = P (t) y(t) (9) i " # N(t) " #" # i yi(t) ! The index i runs over all players! with at least y! oppor- and xD(t) as, " # tunities during year t,and i yi is the total number of opportunities of all N(t)playersduringyeart.Weusea D 1 D P cutoff y 100 which eliminates! statistical fluctuations x (t) = xi (t)= xi(t) ! " # N(t) P (t) N(t) ≡ i i that arise from players with very short seasons. " " # " We now introduce the detrended metric for the accom- x(t) = P " # = P y(t) . (10) plishment of player i in year t, P (t) " # " # P As a result of our detrending method defined by equa- xD(t) x (t) (7) i ≡ i P (t) tion (7), which removes the time-dependent factors that " # Accounting for socio-technological factors that underly achievement A B >

900 D 900 800 800 ≈ 600 pts / season 700 700 600 600 500 500 400 400 300 300 League average < Pts > Basketball

200 Detrended league average < Pts 200 1950 1960 1970 1980 1990 2000 2010 1950 1960 1970 1980 1990 2000 2010 Year statistical baseline Year C 14 D 14 > D 12 12 10 10 ≈ 7 HRs / season 8 8 6 6 4 4 Baseball League average < HR > 2 2 Detrended league average < HR 1920 1940 1960 1980 2000 1920 1940 1960 1980 2000 Year applying Year detrending method

Quantitative measures for success are important for comparing both individual and group accomplishments, often achieved in different time periods.

However, the evolutionary nature of competition results in a non-stationary rate of success, can make comparing accomplishments across time statistically biased. the big debate...Career Home Runs.... 25

Traditional Rank Detrended Rank Rank Name Final Season (L) Career Metric Rank∗(Rank) Name Final Season (L) Career Metric 1BarryBonds2007(22)7621(3) Babe Ruth 1935 (22) 1215 2HankAaron1976(23)7552(23) Mel Ott 1947 (22) 637 3BabeRuth1935(22)7143(26) Lou Gehrig 1939 (17) 635 4WillieMays1973(22)6603(17) Jimmie Foxx 1945 (20) 635 5KenGriﬀey Jr. 2009 (21) 630 5(2) Hank Aaron 1976 (23) 582 6SammySosa2007(18)6096(124) Rogers Hornsby 1937 (23) 528 7FrankRobinson1976(21)586 7(192) Cy Williams 1930 (19) 527 8AlexRodriguez2009(16)583 8(1) Barry Bonds 2007 (22) 502 8MarkMcGwire2001(16)5839(4) Willie Mays 1973 (22) 490 10 Harmon Killebrew 1975 (22) 573 10(18) Ted Williams 1960 (19) 482 11 Rafael Palmeiro 2005 (20) 569 11(13) Reggie Jackson 1987 (21) 478 12 Jim Thome 2009 (19) 564 12(14) Mike Schmidt 1989 (18) 463 13 Reggie Jackson 1987 (21) 563 13(7) Frank Robinson 1976 (21) 444 14 Mike Schmidt 1989 (18) 548 14(10) Harmon Killebrew 1975 (22) 437 15 Manny Ramirez 2009 (17) 546 15(577) Gavvy Cravath 1920 (11) 433 16 Mickey Mantle 1968 (18) 536 16(718) Honus Wagner 1917 (21) 420 17 Jimmie Foxx 1945 (20) 534 17(18) Willie McCovey 1980 (22) 417 18 Ted Williams 1960 (19) 521 18(557) Harry Stovey 1893 (14) 413 18 Frank Thomas 2008 (19) 521 19(5) Ken Griﬀey Jr. 2009 (21) 411 18 Willie McCovey 1980 (22) 521 20(28) Stan Musial 1963 (22) 410

TABLE XII: Ranking of Career Home Runs (1871 - 2009). ...for extensive top-50 tables for Hits, HR, RBI, K, W calculated for single seasons and also over entire the career consult the papers downloadable at:

Methods for detrending success metrics to account for inﬂationary and deﬂationary factors A. M. Petersen, O. Penner, H. E. Stanley. Eur. Phys. J. B 79, 67-78 (2011).

and an analogous statistical analysis of basketball career statistics:

A method for the unbiased comparison of MLB and NBA career statistics across era A. M. Petersen, O. Penner. MIT Sloan Sports Analytics Conference 2012. Physiological/Behavioral components of competition High competition levels can make careers vulnerable to early career negative production shocks (ie stress, burn-out, productivity lulls, etc.)

Achievement-oriented systems: incentives for cut-throat “zero-sum” behavior, i.e. use of performance/cognitive enhancing drugs, possibly leading to blatant cheating/falsiﬁcation

Ethical scandals reveal the price of success

Jan Hendrik Schön Scandal (2001)

On October 31, 2002, Science withdrew eight papers written by Schön On December 20, 2002, Physical Review withdrew six papers On March 5, 2003, Nature withdrew seven papers

Diederik Alexander Stapel Scandal (2011) Social psychologist made up data for at least 30 publications according to preliminary investigation, which is still ongoing. Hisashi Moriguchi Scandal (2012) “Transplant of induced pluripotent stem cells to treat heart failure probably never happened.... He is afﬁliated with University of Tokyo but not with Massachusetts General Hospital nor with Harvard Medical School. The study did not receive Institutional Review Board approval.” nature.com Vol 450|20/27 December 2007 CognizantCOMMENTARY Enhancement Drugs (CED)

Professor’s little helper “ Is it cheating to use cognitive-enhancing The use of cognitive-enhancing drugs by both ill and healthy individuals raises ethical questions that drugs?.... How would you react if you knew NEWS NATURE|Vol 452|10 April 2008 should not be ignored, argue Barbara Sahakian and Sharon Morein-Zamir. your colleagues — or your students — were COMMENTARY NATUREoday there| Volare several 450 drugs|20/27 on the December 2007 market that improve memory, concen- taking cognitive enhancers?... we know that Ttration, planning and reduce impulsive behaviour and risky decision-making, and a number of our scientiﬁc colleagues ...

NEWS manyNATURE more are being|Vol developed. 452|10 Doctors April 2008 NIKOS/ZEFA/CORBIS already prescribe these drugs to treat cogni- already use modaﬁnil [Modiodal, Provigil] 3 tive disabilities and improve quality of life make the transition to independent living . if not encouraged,forPoll patients with results: neuropsychiatric such asdisorders by look air-traffic who’s con- doping to counteract the effects of jetlag, to enhance and brain injury. The prescription use of such Thus, cognitive-enhancing drugs are trollers,drugsIn surgeonsJanuary, is being Nature extended andlaunched to other nurses conditions, an informal who survey work into readers’ long use of cognition-enhancing drugs. Brendan productivity or mental energy, or to deal includingMaher has shift-workers. waded through Meanwhile, the off-label results and found large-scale use and a mix of attitudes towards the drugs. increasingly being considered as possible add- shifts. Oneand non-prescription can even use by imagine the general public situations where with demanding and important intellectual is becominghe US National increasingly Institutes commonplace. of Health is to prescribed for cardiac arrhythmia that also behind ‘other’ which received a few interesting ons to antipsychotic medication, and long- such enhancing-drug-takingAlthoughcrack down the appeal on scientists of pharmaceutical ‘brain doping’ cog- wouldhave an anti-anxiety be recom- effect. Respondents who reasons, such as “party”, “house cleaning” and challenges...” nitiveTwith enhancers performance-enhancing — to help one studydrugs suchlonger, had not taken these drugs, or who had taken “to actually see if there was any validity to the workas Provigil more and effectively Ritalin, a pressor better release manage declared eve- them for a diagnosed medical condition were afore-mentioned article”.

term clinical trials are underway with drugs mended, such as for airport-security screeners, PHOTOTAKE/ALAMY rydaylast week. stresses The release, — is understandable, brainchild of evolution- potentialdirected straight to a simple questionnaire Our question on frequency of use, for those users,ary biologist both healthy Jonathan and Eisen diseased, of the must University“ considerOne about in general ﬁve attitudes. respondents Those who revealed saidwho took they drugs for hadnon-medical used purposes, such as modafinil, which promotes wakeful- or by soldiersof California, Davis, in turnedactive out to be combat. an April that they But had taken there these drugs, are or others, for revealed an even split between those who took 4 the pros and cons of their choices. To enable ness . Although the mechanisms of modafinil no straightforwardthis,Fools’ scientists, prank. And doctors the World and Anti-Brainanswers policy-makersdrugs Dop- andnon-medical, for any non-medical cognition-enhancing fruitful purposes reasons them daily, toweekly, stimulate monthly, or no more than shoulding Authority provide website easy access that it tolinked information to was like- about once a year. Roughly half reported unpleasant thewise advantages fake. But with and a number dangers of of co-conspirators using cognitive- DRUG SOURCES side effects, and some discontinued use because are not fully understood, it has been found debate mustspreading rumours address about receiving each anti-doping situation in turn. of them. Some might expect that negative side enhancing drugs and set out clear guidelinestheir for focus,Answered concentration question: 201 or memory. UseVol 456 |did11 December 2008 affidavits with their first R01 research grants, Skipped question: 1,227 effects would correlate positively with a low to have direct and indirect effects on various theirthe ruse future no use.doubt To gave trigger pause broader to a fewdiscussionnot of the ofdiffer greatly across age-groupsfrequency of use, but ...,that doesn’t which seem to be theserespondents issues we to Natureoffer the’s survey following on readers’ questions, to the case in our sample (see bar graph, below). “...one survey estimated that almost 7% which readers can respond in an online forum. How woulduse of cognition-enhancing you react drugs. if you knewPrescription your Reported side effects included headaches, jit- neurotransmitter systems. Behaviourally, an will surpriseMorning pick-me-up: some. will drugs “ that help you stay alert become as widely acceptable as coffee? Now,The on survey to the was questions. triggered by a Com- (52%) Internet teriness, anxiety and sleeplessness. of students in US universities have used colleaguesmentaryCOMMENTARY by — behavioural or your neuroscientists students — were (34%) Neuroscientist Anjan Chatterjee of the Pollacute results: dose of modafinil has been look found to who’s doping Should adults with severe memory and Barbara Sahakian and Sharon Morein- additional neuro-protective effects. Universitydrug. ofFor Pennsylvania patients with in neuropsychiatric Philadelphia disor- concentrationZamir of the University problems of Cambridge,from neuropsy- Countries withPharmacy ageing populations are seeingpredicts ders, a rise the inbenefits the use of of the these drugs drugs must and be weighed prescription stimulants [Adderall and increase alertness, memory and planning in taking cognitive enhancers? (14%) chiatricUK, who disorders had surveyed be given their colleaguescognitive- a surge in the number of people with Alzheim-other neuroenhancingagainst the potential products short-term and proce- and long-term In January, Nature launched an informal survey into readers’ use of cognition-enhancingenhancingon the use ofdrugs. drugs?drugs that purportedly Brendan enhance er’s. The personal and social costs are stagger-dures asside they effects, become and available these (A. factors Chatterjee should be dis- Ritalin] in this way, and that on some healthy young adults and cognitive flexibility In academia,Wefocus believe and attention the we answer (Nature know is 450,a resounding 1157–1159; that yes. a numbering and in the Unitedof our Kingdom, economicCam. Q.cussed Healthc. with Ethics the individual’s 16, 129–137; doctor 2007). to ascertain 5 A2007).Towards large In debilitating the article, aspectthe two of scientists many responsible neuropsy- asked costs associated with dementias use1 are estimatedofLike cognitive- thethe rise level in cosmetic of acceptable surgery, risk use in of eachcogni- case. campuses, up to 25% of students had in patients with chronic schizophrenia . scientificreaders colleagues whether they would consider in the “boost- Unitedwere asked severalStates additional and questions about tive enhancers is likely to increase as bioethical Maher has waded through the results and found large-scale use and a mix of attitudeschiatric towards disorders is cognitive the impairment. drugs. to rise to £10.9 billion (US$22 billion) by 2031. ing their brain power” with drugs. Spurred by their use. Here’s what they had to say: and psychologicalIf drugs can concerns be shown are overcome to have (see mild side Quick fix: but what are the long-term side effects? Thus,the tremendous cognitive-enhancing response, Nature drugs ran are its a own useful OneAccording in five respondents to a report said commissioned they had used by ‘Worryingthe words’) and as the products gain used them in the past year. These Due to the stated economic and personal the United Kingdom already use modafinil effects, should they be prescribed more therapyinformalenhancing option survey. for1,400 several people disorders, from 60 coun- drugsinclud-drugsAlzheimer’s for non-medical by Research reasonsthe Trust into Cambridge,stimulate healthy UK,cultural acceptance. One difference, Chatterjee costs, the pharmaceutical industry is targeting to counteractingtries Alzheimer’s responded the to disease the online effects and poll. Attention of Deficit jetlag,their treatment focus, to concentration that enhance would or reduce memory. severe Use did cognitive says, is thatwidely use offor cognitive other psychiatric enhancers doesn’t disorders? students are early adopters of a trend he US National Institutes of Health is to prescribed for cardiac arrhythmia that also behind ‘other’ whichHyperactivitySocietyWe asked received must specifically Disorder respond (ADHD).about a threeto few the drugs: growing interesting not impairmentdiffer demand greatly inforacross older cognitive age-groups people by enhancement. (see just line 1% a yearrely on trainingWe That believe responseof medical that cognitive-enhancing specialists must startsuch as by drugs methylphenidateAlzheimer’s disease(Ritalin), isa stimulant a “The nor- chiefgraph, concern left), which willhas surprisebeen estimated some. Nora to cancelsurgeons. with Internet minimal availability side effects will also would greatly also benefit that is likely to grow, and indications drugs that would improve impaired cognition professional. This is particularly important productivityneurodegenerativemallyrejecting used toor thetreat mental idea attention-deficit disease that of ‘enhancement’ energy, hyper- Volkow, or is directorato dirty deal of word, outthe National allwith arguepredicted Institute Henry increases on Greelyaccelerate in andmany use, colleagues heof says.the patients. with schizophrenia, a crack down on scientists ‘brain doping’ have an anti-anxiety effect. Respondents who reasons, such astheactivity “party”, ageing disorder mind but “housecharacter- well-known on cautioningcleaning” college Drug against Abuse and the(NIDA) long-term in Bethesda, care Maryland, costs due to theOur pollcondition found thatfor whichone-third they of are the not drugs yet routinely in specific neuropsychiatric disorders. Often because of potential drug interactions, and so demandingcampuses as and a ‘study aid’;important modafinil (Provigil), intellectualsays that household surveys chal- suggest that stimu-1 being used for non-medical purposes were suggest that they’re not alone.” ized byoday, a decline on university in cognitive campuses use aroundof medications is ageing population . prescribed. Currently, the disorder affects with performance-enhancing drugs such had not taken these drugs, or who had taken “to actually see prescribedif there to treat was sleep disordersany validity but also lant use to is highest the in people aged 18–25 years, purchased over the Internet (see pie chart). The T and behaviouralthe world, functioning,students are striking adverse deals sideto effects.” For all medications, the chief about 24 million people worldwide. when a drugOPINION is approved for one disorder, its we do not advocate self-medication. lenges (seeandused in NATUREoff-label particular graphic to combat learning|Vol general opposite). 456and fatigue|11 December or and Modafinil in students. 2008 concern and cautioning againstrest were obtainedAs with fromAlzheimer’s, pharmacies the or personal on pre- and social overcomeTbuy jet and lag; sell and prescription beta blockers, drugs drugs such as For those who choose to use, methylpheni- scription. It is unclear whether the prescribed as Provigil and Ritalin, a press release declared them for a diagnosed medical condition were afore-mentionedmemory. article”. There are, at present, no treatments their use is adverse side effects that affect the costs are immense, with economic costs in the efficacy in improving cognition in additional other drugsAdderall are and Ritalin available — not to get high,online, but to date but was the their most popular: non- 62% of users for Alzheimer’s disease that can stop or reverse individual’s health and well being. These may United States estimated in the tens of billions of GALLAGHER/SPL C. get higher grades, to provide an edge over their reported taking it. 44% reported taking last week. The release, brainchild of evolution- directed straight to a simple questionnaire Our questionthe onTRENDS decline frequency IN in USE brain OF function, NEUROENHANCERS ofbut cholineste- use, forrange those from mild, temporary physical symp-EFFECT OFdollars DOSE2. It ON is commonSIDE EFFECTS knowledge that people disorders is investigated and thus its use can Would you boost your own brain power? prescriptionfellow students and or to long-term increase in some meas- usemodafinil, has and not 15% saidbeen they had taken raseurable inhibitors80 way their are beingcapacity used for to learning. ameliorate These the betatoms, blockers such suchas dry as propanolol,mouth and reveal- headaches, to withSide schizophrenia effects No sidetypically effects have hallucina- ary biologist Jonathan Eisen of the University about general attitudes. Those who revealed who took drugsimpaired for neural non-medical transmission in the cholin- purposes,more severe side effects such as vomiting and100 tions and delusions, yet it is the long-term cog- be extended to multiple patient groups. In our Cognitive-enhancing drugs are increasingly monitoredtransactions inNever healthy areused crimes in the individuals. United States, ing an overlap between drugs. 80 respond- ergicpunishable system. by Such prison. drugs aim to increase the entsjoint specified pain and other even drugs cardiac that arrhythmiathey were or psy- nitive impairments that often impede everyday 60 80 of California,view, Davis, the turnedoriginal out justification to be an April for drug that treat- they hadbeing taken used these in non-medical drugs, or others, situations for suchrevealed as anFor even many,levels splitMany of acetylcholine, people itbetween seems see such a penaltiesneurotransmitter that those as theappro- who immediatetaking.chosis. took The All mostmedications common and also of these carry was contraindi- function and quality of life for many patients. importantpriate, and for consider maintaining the use ofattention such drugs and to adderall,cations an for amphetamine certain conditions, similar to meth- such as high Even small improvements in cognitive func- ylphenidate. But there were also reports 60 Fools’ prank. And the World Anti-Brain Dop- non-medical, cognition-enhancing purposes them daily, weekly,inbe forming cheating,40 monthly, new unnatural memories, or dangerous. or and no may Yet havemore one blood than pressure, when one should not take the tions could help patients with schizophrenia ment improving quality of life still holds in shift work and by active military personnel. tangible benefits1 of taking these drugs are more survey estimated that almost 7% of students in of centrophenoxine, piractem, dexedrine Used for cognitive enhancement and various alternative medicines such as 40 1157 ing Authority website that it linked to was like- once a year. RoughlyUS universities half have reported used prescription stimu-unpleasant these other disorders. This is where the debate about their use begins persuasive20 than concerns about ginkgolegal and statusomega-3 fatty and acids. lants in this way, and that on some campuses, Adderall is one of several drugs Percentage of respondents of respondents Percentage 20

The most popular reason for taking of respondents Percentage increasingly used to enhance wise fake. But with a number of co-conspirators side effects, and someup to 25%Used discontinuedof for students medical purposes had used them use in the becausethe drugs was to improve concentra- DRUGin earnest. SOURCES How should the use of cognitive- adverse effects. There are clear trendscognitive function. suggest- past year.0 These students are early adopters of tion. Improving focus for a specific task 0 < 25 26–35 36–45 45–55 55–65 > 66 Daily Weekly Monthly Yearly a trend(n =that 330) (isn = likely 487) (n = to 252) grow, (n = 180) and (n = 146) indications (n = 32) (admittedly difficult to distinguish from spreading rumoursDo the aboutsame receivingarguments anti-doping apply for enhancingAnswered question: drugs be 201 regulated in healthyof peo- them. ingSome that might the use expect of stimulants that2 negative such side as methyl- (n = 53) (n = 56) (n = 56) (n = 62) suggest that they’re notAge-group alone . case it would prevent a valid measure of the policy that is neither laissez-faire nor prima- as when a patient says “I know I don’t meetconcentration) diag- ranked a close second Frequency of use affidavits withyoung their children first R01 and research adolescents grants, with ple?Skipped Should question: their 1,227 use always be monitoredeffects by wouldphenidate correlateIn this on article, college wepositively propose campuses actions that with will and campus isa counteracting lowon to divertthe jet to lag rise,enhancement ranked fourth, use. be proven safe and effective, but if one is it will competency of the examinee and would rily legislative. We propose to use a variety of nostic criteria674help society for ADHD, accept the benefits but ofI sometimesenhance- A newer have drug, modafinil (Provigil), has also surely be sought by healthy middle-aged and the ruse noneuropsychiatric doubt gave pause disorders, to a few of such the as those healthcare professionals? frequency of use,ment, but given andappropriatethat is doesn’t becomingresearch and evolved seem more shown to enhancementbecommon- potential. Modafinil is elderly people contending with normal age- therefore be unfair. But if it were to enhance scientific, professional, educational and”Most social would trouble not concentratingregulation. Prescription and drugs arestaying regulated asorganized, approved for the treatment of fatigue caused by related memory decline, as well as by people with ADHD? such not for their enhancing properties but pri- narcolepsy, sleep apnoea and8 shift-work sleep of all ages preparing for academic or licensure respondents to Naturelong-term’s survey learning, on readers’ we may be more willing If offeredresources, by in a additionfriend or to col-legislation, to shapethe case and in ourit would samplemarily help for considerationsplace (seeme to barin ofhave safety ever graph, and some potential younger Ritalin below).disorder. students Iton is currently prescribed. off label for a examinations. to accept enhancement. After all, unlike ath- a rational, evidence-based policy informedconsider handthat for an daysabuse. when Still, cognitive I really enhancement need has to much be onwide top range of neuropsychiatric and other medi- use of cognition-enhancingAt present, children drugs. diagnosed with ADHD Prescriptionleague, would you, the reader, Reported side effectsto offer individualsUniversities included and society, andheadaches, amay proper havecal conditions jit- to decideinvolving fatigue 5 as well as for Favouring innovation letic competitions, in many cases cognitive by a wide array of relevant experts and stake- things at work.”societal response Physicians will involve makingwho enhance- view medicinehealthy people who need to stay alert and awake Human ingenuity has given us means of enhanc- The surveyare was routinely triggered prescribed by a Com- long-term medi- (52%)take a pill that would helpInternet you espressoteriness, confers anxietyments and availablewhether sleeplessness. while managing to their banrisks. drugwhen sleep use deprived, alto- such as physicians on night ing our brains through inventions such as writ- cations includingenhancements atomoxetine are not zero-sum and stimu- games. Cog-to betterholders. focus, Specifically, plan or remem- we(34%) propose four types of as devoted to healinggether, will view or to such tolerate prescribingcall6. In itaddition, in laboratorysome studies have shown ten language, printing and the Internet. Most mentary by behavioural neuroscientists an unfairNeuroscientist advantagePaths to enhancementAnjan Chatterjee thatof modafinil the enhances aspects of executive authors of this Commentary are teachers and nitive enhancement, unlike enhancement for policy mechanism. as inappropriate,Many of the whereas medications usedthose to treat who psychi- viewfunction medi- in rested healthy adults, particularly strive to enhance the minds of their students, Barbara Sahakianlants, such andsports as Sharon methylphenidate competitions, Morein- could (Ritalin) lead to substantive and ber? UnderThe first what mechanism conditions is an acceleratedUniversity cine more of broadlyPennsylvaniaatric and neurologicalsituations as helping conditions in patients(whether also Philadelphia improve liveinhibitory to better enable control 7. Unlike all- Adderall and Rita- both by adding substantive information and by at work.” the performance of the healthy. The drugs most lin, however, modafinil prescriptions are not showing them new and better ways to process Zamir of theamphetamine. Universityimprovements of Both Cambridge, methylphenidate in the world. and wouldprogramme youPharmacy feel of comfortable research to build a knowledgepredicts or achieve a rise commonly theirin the goals usednight use for would cognitive of study these enhancement be opensessions drugs at tocommon, consid- andor and to the boost drug is consequently rare on that information. And we are all aware of the present are stimulants,22 namely Ritalin (methy- the college black market. But anecdotal evidence abilities to enhance our brains with adequate UK, who hadatomoxetine surveyedFairness increasetheir colleagues in the cognitive levels enhancementsof the neuro- hastaking a base a pill, concerning(14%) and under the usage,what conditions benefits and would asso-other neuroenhancingeringalertness such phenidate)a requestduring and Adderall productslectures).. There (mixed isamphetamine certainly and proce-and a a prec- readers’ survey both suggest that adults exercise, nutrition and sleep. The drugs just salts), and are prescribed mainly for the treat- sometimes obtain modafinil from their physi- reviewed, along with newer technologies such dimension beyond the individual. If cognitive ciated risks of cognitive enhancements. Good edent for this broader view in certain branches 2 on the use oftransmitter drugs that noradrenaline. purportedly enhance Generally, the thera- you decline? dures as theyThe become debatement of attention overavailable deficit cognitive-enhancing hyperactivity (A. disorder Chatterjee cians or online for drugs enhancement purposes . as brain stimulation and prosthetic brain chips, enhancements are costly, they may become the policy is based on good information, and there of medicine,(ADHD). including Because of theirplastic effects onsurgery, the cat- derma-A modest degree of memory enhancement should be viewed in the same general category peutic effects of these drugs include reductions The answers to such questions hinge on must alsoecholamine consider system, these the drugs expected increase exec- is magnitude possible with the ADHD of medications just as education, good health habits, and informa- focus and attention (Natureprovince 450, of the 1157–1159; rich, adding to the educational is currently much we do not know aboutCam. the Q.tology, Healthc. sportsutive Ethics functionsmedicine in 16,patients and 129–137; and fertility most healthy medicine. 2007).mentioned as well as with medications devel- tion technology — ways that our uniquely inno- in inattention, hyperactivity and impulsivity, many factors, including the exact drug being the benefitsnormal people, and improving weigh their them abilities toagainst oped for the the treatment risks of Alzheimer’s disease vative species tries to improve itself. 2007). In the article, theadvantages two scientists they already asked enjoy. One could miti- short- and long-term benefits and risks ofLike the the riseBecause in cosmeticfocus physicians their attention, surgery, manipulate are the information use gatekeepers of cogni-such as Aricept to (donepezil), which raise levels Of course, no two enhancements are equiva- although their widespread and long-term use discussed, its short-term and long-term ben- and sidein workingeffects memory of and eachflexibly control drug. their Mostof acetylcholine readers in the brain 8. Several other lent in every way, and some of the differences readers whether they gatewould this consider inequity by “boost- giving everywere exam-taker asked several cognitive-enhancement additional questions drugs about currently beingtive enhancers medications isresponses likely discussed3. These to drugsincrease arehere, widely usedsociety as thera- bioethical lookscompounds to with different pharmacological have moral relevance. For example, the ben- in youngerfree children access tohas cognitive been controversial. enhancements, as someefits used,and risks, and about and who the is purpose using them for and which why. Forit is themwould for guidancenotpeutically. consider With on rates theof ADHD that use in theofhaving range these of medica-actionsa double are in early clinicalshot trials, having shown efits of education require some effort at self- ing their brain power” with drugs. Spurred by their use. Here’s what they had to say: and psychological4–7% concerns among US college arestudents overcome using DSM positive (see effects on memory in healthy research improvement whereas the benefits of sleep do ADHD isschools a heritable provide and computers disabling during condition exam weekused. example, There are what instances are the patterns in which of use most outside people of tionsof espresso and devices,criteria 4or, and astimulant and soft guidelinesmedication drink the containing stand- fromsubjects other (see, caffeine for example, ref. 9). It is too early not. Enhancing by nutrition involves changing the tremendous response, Nature ran its own One in five respondents said they had used ‘Worrying words’)ard therapy, and there asare plenty the of theseproducts drugs on to knowgain whether any of these new drugs will what we ingest and is therefore invasive in a way characterizedto all bystudents. core cognitiveThis would andhelp levelbehav- the play-wouldthe agree United that States the and use outside of cognitive-enhanc- of college commu- professionalwould confer702 groups an unfairwill need advantage to take into at work. informal survey.ioural 1,400 symptomsing people field. of fromimpulsivity, 60 coun- hyperactivitydrugs for ingnon-medical drugsnities? Whatshould reasons are thebe risks prevented to stimulateof dependence or at when culturalleast account acceptance.The use the of gatekeepers’ caffeine One difference, to policies. enhance Chatterjee For concentrationthis rea- tries respondedand/or to theinattention. onlinePolicy poll. governing ADHD affects the use 4–10% theirof cognitive offocus, regulated concentration used for and cognitive monitored, or memory. enhancement? such Use as didWhat by healthy specialsays, is thatson,is commonplace, theuse responsibilities of cognitive despite that enhancers physicians having doesn’t bearside foreffects in We askedchildren specifically worldwide,enhancement about and three in iscompetitive the drugs: most situationsprevalentnot differ should greatly risks across arise with age-groups the enhancement (see line of children’s rely on thetrainingat consequences least someof medical individualsof their specialists decisions9. Often are such particu- overlooked as methylphenidateneuropsychiatric (Ritalin),avoid exacerbating a disorder stimulant of socioeconomic childhood.nor- graph, inequali- left), whichcognition? will Howsurprise big are some. the effects Nora of currentlysurgeons. larlyin Internet media sensitive, reports availability being effectively on cognitive will decisions also enhancers greatly for all is the mally used to treat attention-deficitties, and should take hyper- into accountVolkow, the validity director available of the enhancers? National Do Institute they change on ‘cogni-accelerateof us. use, It would he says. therefore be helpful if physicians ADHD is associatedof enhanced with test increasedperformance. lev- In developing tive style’, as well as increasing how quickly asfact a profession that many gave of theserious effects consideration in healthy to individ- activity disorderels of drop-outs but policywell-known from for this education, purpose, on college problemsjob Drug of enforce- Abuse (NIDA)and accurately in Bethesda, we think? Maryland, And given that mostOur pollthe ethics found ofuals appropriatethat are one-third transient prescribing ofand the small-to-mod- of drugs cogni- campuses asdismissal, a ‘study aid’; criminalment modafinil must activities, also be (Provigil), considered. sub- Insays spite that of strin- householdresearch surveys so far has suggest focused that on simplestimu- laboratorybeing usedtive enhancers, for non-medical anderate consulted in size.purposes widely Just as as towere one how would prescribed stanceto treat abuse, sleepgent other regulation, disorders mental athletes illnessbut also continue lant to use, use and is be highest tasks, in how people do they aged affect 18–25 cognition years, in thepurchased real to strike over the the balance Internethardly of limits (see propose forpie patient chart). that benefit aThe strong cup used off-labeland to accidents combatcaught6. Long-termgeneral using, banned fatigue drug performance-enhancing or and in students. world? Do they increase the total knowledgerest were and obtained protection from in ofa pharmaciesliberal coffee democracy. could or be on Exam-the pre- secret of overcome jettreatment lag; and drugs.seems beta to blockers, be beneficial drugs For those whoand choose understanding to use, methylpheni- that students take scription.with ples It of is such unclear limits inwhetheracademic other areas the achievement of prescribedenhancement or faster We call for enforceable policies concern- them from a course? How do they affect various medicine include the psychological screening of in many cases. date was the most popular: 62% of users career advancement, the use of

ing the use of cognitive-enhancing drugs to aspects of occupational performance? candidates for cosmetic surgery or tubal ligation, SEAGRAVES/PHOTONICA/GETTY R. However, the side effects of chronic such drugs does not necessarily TRENDS IN USE OFsupport NEUROENHANCERS fairness, protect individualsreported from takingWe it.call 44% for a reported programme taking of research intoEFFECT the andOF upperDOSE bounds ON SIDE on maternal EFFECTS age or number drug use maycoercion only and become minimize noticeable enhancement-related in modafinil, anduse and 15% impacts said they of cognitive-enhancing had taken drugs of embryos transferredentail cheating. in fertility treatments. 80 the longersocioeconomic term, for example, disparities. with apparentbeta blockersby such healthy as propanolol,individuals. reveal- TheseSide examples effects of Cognitive limits No may side not effects enhancersbe specified by with The second mechanism is the participa-100 law, but rather by professional standards. Neverreductions used in normal growth rates in chil-ing an overlap between drugs. 80 respond- small or no side effects but with moder- dren with MaximumADHD who benefit, are taking minimum stimulant harm tion of relevant professional organizations Otherate professionalenhancing groupseffects to that which alleviate this for- The7 new methods of cognitive enhance-ents specified otherin drugs formulating that they guidelines were for their recommendation applies include educators 60 medication . In fact, for many drugs there is 80 getfulness or enable one to focus better on ment are ‘disruptive technologies’ taking. The most commonmembers of these in relation was to cognitive and human-resource professionals. In differ- SPL limited information on long-term effects and Is it cheating to use cognitive-enhancing drugs? the task at hand during a tiring day at work that could have a profound adderall,7 an amphetamine similarenhancement. to meth- Many dif- ent ways, each of these professions has respon- in many areaseffect the on findingshuman life are in inconsistent the ylphenidate.. But there were alsoferent reports professions have60 a sibilitywould for be fostering unlikely and to evaluating meet much cognitive objection. 40 Consequently, in all the cases outlined above, children or in competitive settings (including And does it matter if it is delivered as a pill twenty-first century. A of centrophenoxine, piractem,role dexedrine in dispensing, using performance and for advising individuals who we believelaissez-faire the medical approach consensus would be entrance exams to university).or working with peo- areor seeking a drink? to improve Would their you, performance, the reader, and welcome Used for cognitive enhancement and various alternative medicines such as 40 that medicationto these choice,methods dose will and timing, There are also situationsple who in whichuse cognitive many somea cognitive responsibility enhancer also for delivered protecting in the a bever- 20 therapeuticleave effects us andat the safety mercy should be moni-ginkgo andwould omega-3 agree that fatty the acids. use enhancers.of drugs to By improve creating interestsage that of thoseis readily in their obtainable charge. In contrast and afford-

Percentage of respondents of respondents Percentage of powerful market policy at the level 20of to physicians, these professionals have direct tored for individual patients by a healthcareThe mostconcentration popular orreason planning for taking may be tolerated, of respondents Percentage able, and has a moderate yet noticeable effect Used for medicalforces purposes that are bound the drugs was to improve concentra-professional societies, conflicts of interest that must be addressed in to be unleashed by the it will be informed by whatever guidelines they recommend: liberal 0 1158 tion. Improving focus for a specific task 0 < 25 26–35 36–45promise 45–55 of increased55–65 > 66 the expertise of these useDaily of cognitiveWeekly enhancers Monthly wouldYearly be expected (n = 330) (n = 487) (nproductivity = 252) (n = 180) and(n = 146)competi- (n = 32) (admittedly difficult to distinguishprofessionals, from and their to(n =encourage 53) (n = 56)classroom (n = 56) order and(n = 62) raise stand- Age-group tive advantage. The concerns concentration) ranked a closecommitment second to the goals of ardized measuresFrequency of student of use achievement, both about safety, freedom and fair- and counteracting jet lagtheir ranked profession. fourth, of which are in the interests of schools; it would ness, just reviewed, may well One group to which this also be expected to promote workplace produc- 674 seem less important than the The prescription drug Ritalin is recommendation applies is tivity, which is in the interests of employers. attractions of enhancement, illegally traded among students. physicians, particularly in Educators, academic admissions officers and for sellers and users alike. primary care, paediatrics and credentials evaluators are normally responsible Motivated by some of the same considera- psychiatry, who are most likely to be asked for for ensuring the validity and integrity of their tions, Fukuyama21 has proposed the formation cognitive enhancers. These physicians are some- examinations, and should be tasked with for- of new laws and regulatory structures to protect times asked to prescribe for enhancement by mulating policies concerning enhancement by against the harms of unrestrained biotechno- patients who exaggerate or fabricate symptoms test-takers. Laws pertaining to testing accom- logical enhancement. In contrast, we suggest a of ADHD, but they also receive frank requests, modations for people with disabilities provide 704 Ethics and scientific careers • Competition (“fairness”): • strategizing / extreme behavior, e.g. scientific fraud • CED (cognitive enhancing drugs) free-riding in team science, individual vs group: • the“tragedy of the scientific commons” • Funding: • financial incentives & who should subsidize early career risk • how to attribute / appraise / reward achievement, especially in the case of extremely large team projects • Careers: predicting future career achievement using incomplete information and poorly understood/ designed achievement measures Food for thought

• Competition and Reward: There are many analogies between the superstars in science and the superstars in professional sports, possibly arising from the generic aspects of competition. Currently, the contract length, compensation, and appraisal timescale in these two professions are VERY different. Is science becoming more like professional sports? • Science as an evolving institution: An institutional setting that neglects specific features of academic career trajectories (increasing returns from knowledge spillovers and cumulative advantage, collaboration factors, career uncertainty) is likely to be inefficient and unfair. But what is “fair”? • Complex career dynamics: Knowledge, reputation, and collaboration spillovers are major factors leading to increasing returns along the scientific career trajectory. A data-centric (“big data”) understanding of the production function of individual scientists can improve academic policies aimed at increasing career sustainability by decreasing career risk.

• Quantitative and empirical demonstration of the Matthew effect in a study of career longevity, A. M. Petersen, W.-S. Jung, J.-S. Yang, H. E. Stanley. Proc. Natl. Acad. Sci. USA 108, 18-23 (2011). Thank You! • Persistence and Uncertainty in the Academic Career, A. M. Petersen, M. Riccaboni, H. E. Stanley, F. Pammolli. Proc. Natl. Acad. Sci. USA 109, 5213-5218 (2012). A special thanks to my collaborators: • On the distribution of career longevity and the evolution of home run prowess in professional Santo Fortunato, Woo-Sung Jung, baseball, A. M. Petersen, W.-S. Jung, H. E. Stanley, Europhysics Letters 83, 50010 (2008). Fabio Pammolli, Raj Pan, Orion • Methods for measuring the citations and productivity of scientists across time and discipline, A. M. Penner, Massimo Riccaboni, Gene Petersen, F. Wang, H. E. Stanley. Phys. Rev. E 81, 036114 (2010). Stanley, Sauro Succi, Fengzhong • Methods for detrending success metrics to account for inflationary and deflationary factors, A. M. Wang, and Jae-Sook Yang Petersen, O. Penner, H. E. Stanley , Eur. Phys. J. B 79, 67-78 (2011). http://physics.bu.edu/~amp17/ • Reputation and impact in academic careers, A. M. Petersen, S. Fortunato, R. K. Pan, K. Kaski, O. Penner, M. Riccaboni, H. E. Stanley, F. Pammolli, (submitted) ArXiv:1303.7274 Ascent in competitive arenas: Abstract: From Fenway Park to Mass Ave Competitive arenas are abundant in society and are characterized by at least three basic principles: limited opportunities, cumulative advantage, and the boundless ambitions of highly driven individuals. Using longitudinal career data for several hundred top-cited physicists, biologists, and mathematicians, I will show that stellar careers can be classified by common growth patterns. And while much is known about the stellar ascent of the likes of Mozart, Babe Ruth, and Einstein, little is known about their numerous out-shined competitors. Using data from six high- impact journals complemented by comprehensive career data spanning the entire history of the Major League Baseball labor force, I will further illustrate how the skewed distributions for diverse career achievement measures can be explained by simple models for career progress and competition. Context also matters, and one cannot understate the role that institutions play in establishing competitive norms and terms of fair play. As science continues to evolve towards a bigger and more interconnected system, an institutional setting which neglects the features of competition may inadvertently give rise to shifts in performance incentives and promote a “tragedy of the scientific commons” marked by the dilemma of individual versus the group. To this end, there is an increasing need to better understand the ethics of competition, as evidenced by both the frequency of research scandals and the widespread emergence of performance- (and even cognitive-) enhancing drugs in society’s competitive arenas, which together highlight the risk that individuals are willing to accept in their pursuit for even the slightest competitive advantage.