The Anna Karenina Principle: A Way of Thinking About Success in Science
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
The Anna Karenina Principle: A Way of Thinking About Success in Science Lutz Bornmann Division for Science and Innovation Studies, Administrative Headquarters of the Max Planck Society, Hofgartenstraße 8, D-80539 Munich, Germany. E-mail: bornmann@gv.mpg.de Werner Marx Max Planck Institute for Solid State Research, Heisenbergstraße 1, D-70569 Stuttgart, Germany. E-mail: w.marx@fkf.mpg.de The first sentence of Leo Tolstoy’s (1875–1877/2001) number of things required for a happy family and an novel Anna Karenina is: “Happy families are all alike; unhappy family lacks a certain constellation of aspects (or at every unhappy family is unhappy in its own way.” Here, least one aspect), each unhappy family is unhappy in its own Tolstoy means that for a family to be happy, several key aspects must be given (e.g., good health of all family very specific way. members, acceptable financial security, and mutual Diamond (1994, 1997) extended the principle behind affection). If there is a deficiency in any one or more of Tolstoy’s first sentence of the novel to understanding the these key aspects, the family will be unhappy. In this requirements for success in complex undertakings, calling it article, we introduce the Anna Karenina principle as a the Anna Karenina principle (AKP). According to the AKP, way of thinking about success in science in three central areas in (modern) science: (a) peer review of research for something to succeed, several key aspects or conditions grant proposals and manuscripts (money and journal must be fulfilled. Failure in any one of these aspects leads to space as scarce resources), (b) citation of publications failure of the undertaking. That is, the success of complex (reception as a scarce resource), and (c) new scientific undertakings always depends upon many factors, each of discoveries (recognition as a scarce resource). If which is essential; if just one factor is lacking, the undertak- resources are scarce at the highly competitive research front (journal space, funds, reception, and recognition), ing is doomed. The AKP is related to the conjunctive deci- there can be success only when several key prerequi- sion rule in consumer behavior research or psychology sites for the allocation of the resources are fulfilled. If (Gilbride & Allenby, 2004), by which in a decision situation any one of these prerequisites is not fulfilled, the grant a person selects only those objects (the successful objects) proposal, manuscript submission, the published paper, that are found acceptable on all relevant criteria. (A different or the discovery will not be successful. decision rule is, for example, a compensatory rule: For an object to be successful, a negative or unacceptable rating on Introduction one criterion can be compensated by a positive or acceptable rating on another criterion.) The first sentence of Leo Tolstoy’s (1875–1877/2001) Based on the AKP as stated by Diamond (1994, 1997), novel Anna Karenina is: “Happy families are all alike; every there are important implications for the conditions of unhappy family is unhappy in its own way” (p. 1). Here, success and failure: (a) Even though “we tend to seek easy, Tolstoy means that for a family to be happy, several key single-factor explanations for success” (Diamond, 1994, aspects must be given (e.g., good health of all family p. 4), there are usually not any for complex undertakings; members, acceptable financial security, and mutual affec- (b) “success actually requires avoiding many separate causes tion). If there is a deficiency or failure in any one or more of of failure” (Diamond, 1997, p. 157), and if only one cause of these key aspects, the family will be unhappy. As there are a failure is avoided, there will be no success; (c) “no one property guarantees success, but many can lead to failure” Received November 23, 2011; revised January 25, 2012; accepted January (Shugan, 2007, p. 145); (d) as favorable outcomes require 26, 2012 every detail to be right whereas an unfavorable outcome © 2012 ASIS&T • Published online 10 September 2012 in Wiley Online only requires one wrong detail; (e) “favorable outcomes are Library (wileyonlinelibrary.com). DOI: 10.1002/asi.22661 rare and more informative than unfavorable outcomes” JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY, 63(10):2037–2051, 2012
(Shugan & Mitra, 2009, p. 11). However, this holds only for several key prerequisites for the allocation of the resources adverse environments: “In propitious environments, favor- are fulfilled. If any one of these prerequisites is not fulfilled, able outcomes convey less information (Shugan & Mitra, the grant proposal, the manuscript submission, the published 2009, p. 13). paper, or the discovery is doomed to fail in the highly com- The AKP has been applied in many fields of research in petitive environment at the excellence level. recent years. There is widespread use of the AKP in history The AKP can yield important insights for understanding and ecology, among others. Diamond (1994, 1997) used it to success in science at the excellence level because it contra- find answers to the question: “Why have so many seemingly dicts common assumptions. We frequently associate success suitable, big, wild mammal species, such as zebras and with special characteristics of the successful persons and peccaries, never been domesticated, and why were the emphasize what is unusual about them (e.g., see the inter- successful domesticates almost exclusively Eurasian?” views with highly cited scientists at Thomson Reuters’ (Diamond, 1994, p. 4). In accordance with the AKP, Dia- ScienceWatch.com). When examining success, we mostly mond’s (1994, 1997) theory names several necessary con- do not think about a lack of success (science at most at ditions related to geography for successful domestication the medium level). The AKP changes the focus on the (e.g., quick growth rate, no nasty disposition with tendency nonsuccessful as an individual case that through its unique- to kill humans). Moore (2001) applied the AKP to ecological ness is not successful. A successful thing is the standardized risk assessments: “Following from the Anna Karenina prin- thing, which fulfills all requirements. In this way, for ciple, there are many ways to ruin an ecological risk assess- instance, the AKP could explain the tendency in peer review ment, but only a few pathways to success” (p. 236). McClay for making conservative decisions (Lamont, 2009). In the and Balciunas (2005) applied the principle to the area of selection of grant recipients, a grant proposal is success- biological control of weeds. In the field of empirical mar- ful only if all of a prestigious funding organization’s pre- keting research, Shugan (2007) formulated the “Anna determined criteria are fulfilled; a risky, novel research Karenina bias” as follows: approach outlined in a grant proposal is not successful in peer review in its own way because it does not fulfill all If we only observe survivors and survivors share the critical criteria that the mainstream in a research area considers properties necessary for survival, then there will be little or no important. variation on the key variables (or constants) related to these The AKP has a general quality that enables it to be used properties. Hence, it will be difficult to infer the descriptive in combination with theories. For example, the AKP will be theory leading to success from the passive observation of used later as a broader context for Merton’s (1973) socio- survivors. We would need to actively observe nonsurvivors. logical theory of science and the constructivist sociology of (p. 146) science. We see the AKP as a methodological principle to study success in competitive situations rather than as a sub- Shugan and Mitra (2009) applied the AKP in an article on stantive theory of success in science. It is a certain way of using averaging and nonaveraging statistics for success thinking about prerequisites of success in science at the metrics: “When environments are adverse (e.g., failure- excellence level. It is not our intention in this work to show rich), nonaveraging metrics correctly overweight favorable precisely which given set of prerequisites is necessary and outcomes. We refer to this environmental effect as the Anna sufficient to be successful in peer review of research grant Karenina effect, which occurs when less favorable outcomes proposals and manuscripts, citation of publications, and new convey less information” (p. 4). scientific discoveries. This work introduces AKP to science In this article, we use the AKP as a way of thinking about studies only and describes some possible prerequisites for success in science. Those researchers in science studies that success in the three areas. Thus, AKP can be seen as a also are active in the history of science will often have read hypothesis that can be empirically tested in quantitative Diamond’s (1994, 1997) study. In this article, we focus on science studies. It should be the task of future empirical science at the excellence level (Bornmann, de Moya- research to identify for each area, as precisely as possible, Anegón, & Leydesdorff, 2010). Since resources are scarce the set of prerequisites. here, it is exceedingly difficult to be successful. We address the effect of AKP in three central areas in modern science: Peer Review and the AKP (a) peer review of research grant proposals and manuscripts (money from prestigious funders and space in high-impact Aside from the selection of manuscripts for publication journals as scarce resources), (b) frequent citing of publica- in journals, the most common contemporary application of tions (reception as a scarce resource), and (c) discoveries peer review in scientific research is for the selection of leading to scientific breakthroughs (recognition as a scarce fellowship and grant applications. Peers or colleagues asked resource). This article draws attention to looking at success to evaluate applications or manuscripts in a peer-review in science as the outcome of a number of interacting factors process take on responsibility for assuring high standards in rather than as a one-dimensional affair. If resources are various research disciplines. According to Lamont (2009), scarce at the excellence level (journal space, funds, recep- “peers monitor the flow of people and ideas through the tion, and recognition), there can be success only when various gates of the academic community” (p. 2). Although 2038 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—October 2012 DOI: 10.1002/asi
equals active in the same field may have limited vision due external reviews, a manuscript must be reviewed positively to their membership in the specialist group, they “are said to by both reviewers to be accepted for publication, and if three be in the best position to know whether quality standards reviews toward a decision are available to the editors, at least have been met and a contribution to knowledge made” two must be positive. (Eisenhart, 2002, p. 241). Peer evaluation in research thus Second, Bornmann and Daniel (2005) investigated the entails a process by which a selective jury of equals active in relationship of selection criteria and decisions in the com- a given scientific field convenes to evaluate the undertaking mittee peer-review process used by an international founda- of scientific activity or its outcomes. The jury of equals may tion for the promotion of basic research in biomedicine, the be consulted as a group or individually, without the need for Boehringer Ingelheim Fonds (BIF), for awarding long-term personal contacts among the evaluators. The peer-review fellowships to postgraduate researchers. For approval or process lets the active producers of science, the experts, rejection of fellowship applications, three criteria are deci- become the “gatekeepers” of science (McClellan, 2003). sive, according to Hermann Fröhlich (2001), former manag- Other than the studies on reliability (i.e., is the selection ing director of the BIF: “In addition to the applicant’s [track] of manuscripts or grant proposals reliable, or is it a chance record and the originality of the research project, there is a result?), fairness (i.e., are certain groups of authors or appli- third element on which our judgment is based: the quality of cants favored or at a disadvantage?), and predictive validity the laboratory in which the applicant wants to pursue his (i.e., does the process fulfill the objective to select the best project” (p. 73). Bornmann and Daniel (2005) tested the manuscripts or proposals?) (Bornmann, 2011), there have extent to which approval of an application depends funda- been few studies in the peer-review research literature on the mentally on positive assessments on all three criteria using evaluation process as it occurs. Few studies have dealt with the Boolean probit statistical technique. This statistical pro- the typical review criteria of editors or program managers cedure introduced by Braumoeller (2003) allows binary out- and with the editors’ or program managers’ rules for deci- comes (here, 0 = rejection, 1 = approval) to be modeled as sion making on these criteria that effectively yield the out- the results of Boolean interactions among independent comes of the peer-review process. Although many journals causal processes. The results using this procedure confirm a and funding agencies name a number of criteria that they use conjunctural causation of the approval decision. In agree- in their peer-review process (for an overview, see Lamont, ment with the prescriptive principles of the BIF (Fröhlich, 2009), it is still unclear what criteria are in fact decisive for 2001), a positive assessment of the research project, in con- success and failure (see Thorngate, Dawes, & Foddy, 2009). junction with both a positive assessment of the laboratory in Sonnert (1995) indicated that “in stark contrast with the question and a positive assessment of the applicant’s multi-faceted relevance of peer review in science, the peer achievement record, proves to have a highly statistically review process has largely retained the characteristics of a significant effect on approval of the application. ‘black box’. It does produce quality judgments, but one does The findings of Bornmann and Daniel (2009, 2010) on not quite know how they come about” (pp. 37–38). Gosden AC-IE and Bornmann and Daniel (2005) on BIF show (2003) explained the shortage of studies as follows: “As clearly that for success in the peer-review process, for the gatekeeping discourse, peer reviews remain largely under- reviewers and decision-makers all criteria must be fulfilled. researched principally due to their hidden status and issues If reviewers at AC-IE assess a submitted manuscript as unim- of confidentiality” (p. 87). portant or if the BIF board of trustees rates the laboratory In the following, we look at two studies published in where the grant applicant intends to conduct the research as recent years that dealt with the criteria or decision rules in not very favorable, they do not accept the manuscript for peer review. The findings of both studies have indicated that publication or approve the grant. In agreement with the AKP, for success in the review process, the AKP plays a decisive for success in the peer-review process all factors must be role. given; just one missing factor will lead to failure. First, taking the journal Angewandte Chemie Interna- Through the increasing shift toward the “soft-money tional Edition (AC-IE) as an example, Bornmann and Daniel system” in the financing of research and the use of papers as (2009, 2010) investigated the editors’ rules for decisions on an evaluative measure of scientific performance (published acceptance or rejection of manuscripts as the outcome of the by individual researchers or groups of scientists) in nearly peer-review process. Interestingly, the editors at AC-IE all scientific disciplines, the peer-review processes by follow a “clear-cut” rule in most cases: The editors accept research funding organizations and journals are confronted for publication only those manuscripts that the reviewers with an ever-increasing number of submissions (e.g., Gölitz, have evaluated positively regarding importance and print- 2008). This development poses new challenges to the pro- worthiness in the AC-IE. Thus, a manuscript is published cesses. In earlier days, reviewers needed to reliably filter out only if two reviewers choose the response category “very proposals or manuscripts that did not meet a certain important” or “important” in answer to the question “how minimum standard (negative selection). In those days, the important do you consider the results?” and also do not AKP most probably did not play much of a role in success in answer “no” to the question “do you recommend acceptance the peer-review process: Since the resources were not too of the Communication?” The editors appear to deviate only scarce, not all factors for the success of a grant proposal or rarely from this rule. If the editors decide on the basis of two submitted manuscript had to be given. Today, however, JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—October 2012 2039 DOI: 10.1002/asi
reviewers usually need to select the “best” from a multitude (2008) collected data on the publication output of 758 Cana- of high-quality scientific papers and proposals (positive dian political scientists and found that: selection). According to Yalow (1982), for today’s peer review the question is “how to identify the few, those who [a]n article is more likely to be widely cited if it is published in make the breakthroughs which permit new horizons to open, a prestigious journal, if it is written by several authors, if it from the many who attempt to build on the breakthroughs— applies quantitative methods, if it compares countries, and if it often without imagination and innovation” (p. 401). And deals with administration and public policy or elections and with these few, in accordance with the AKP, all factors must political parties. (p. 802) be given to be successful in today’s highly competitive research environments. Lansingh and Carter (2009) identified the following vari- ables as having the greatest effect on citation counts: number of authors, country/region of publication, subject area, lan- Citations and the AKP guage, and funding. In a very large-scale study, Vieira and The aim of scientists publishing a paper is not only to Gomes (2010) looked at the whole set of the more than present research findings to fellow researchers and to put 220,000 articles that were published in 2004 and referenced their names on the findings (Merton, 1957) but also to in Thomson Reuters’ Web of Science for Biology & Bio- invite fellow researchers “to freely take and use the infor- chemistry, Chemistry, Mathematics, and Physics. These mation it [the paper] contains” (McClellan, 2003, p. 41). researchers concluded that “[t]he number of co-authors, the It complies with the rules of good scientific practice for number of institutional addresses, the number of pages, the scientists to accept the invitation and to relate their own number of references and the journal impact factor were perspectives, significant concepts, chosen methodologies, considered as basic features that may have direct influence and definitions of problems to peers’ perspectives, con- on the citation count” (p. 11). cepts, methodologies, and definitions. The method of indi- The results of these and several other studies (for an cating these relationships is through references or citations overview, see Bornmann, Mutz, Neuhaus, & Daniel, 2008) (Hooten, 1991). References link “documents and authors have indicated that the research activities of scientists, pub- in accordance with the commonly perceived dynamics lication of their findings, and citation of the publications by of knowledge production” (de Bellis, 2009, p. 14). In the colleagues in the field also are social activities. This means area of research evaluation, if a paper has a high cita- that citation counts for the scientists’ publications are not tion impact, a great many citing papers have been “built only an indicator of the impact of their scientific work on the upon” this cited paper. These highly cited papers are called advancement of scientific knowledge but also the result of “hot papers,” “fast breaking papers,” or—if the citation many social factors besides the scientific quality of the sci- impact sets in only after a certain period of time—”sleep- entists’ research. Against this backdrop, two competing ing beauties.” theories of citing have been developed in past decades, both However, the usefulness of citation counts for measuring of them situated within broader social theories of science. research impact has been questioned (see Joint Committee One is often called the normative theory of citing and the on Quantitative Assessment of Research, 2008). In 1972, other the social constructivist view of citing. Both of these Garfield, founder of the Institute for Scientific Information citation theories attempt to clarify the fundamental question and its chairman emeritus, noted: as to why author x cited article a at time t (Sandström & Sandström, 2008). Citation frequency is . . . a function of many variables besides The normative theory, following Merton’s (1973) socio- scientific merit: an author’s reputation, controversiality of logical theory of science, basically states that scientists give subject matter, circulation, availability and extent of library credit to colleagues whose work they use by citing that holdings, reprint dissemination, coverage by secondary ser- work. Thus, citations represent intellectual or cognitive vices, priority in allocation of research funds, and others. It is influences on scientific work. Merton (1988) expressed this extremely difficult, even when possible, to clarify the relations among such variables and their relative impact on citation fre- aspect as follows: quency. (p. 476) The reference serves both instrumental and symbolic functions in the transmission and enlargement of knowledge. Instrumen- According to Gölitz (2005), other factors affecting tally, it tells us of work we may not have known before, some citation counts are “the number of researchers currently of which may hold further interest for us; symbolically, it reg- working on that topic; articles on unfashionable or highly isters in the enduring archives the intellectual property of the specialized topics, which can certainly be or become very acknowledged source by providing a pellet of peer recognition important, are naturally less cited than articles on current, of the knowledge claim, accepted or expressly rejected, that main-stream research” (p. 5539). was made in that source. (p. 622) (also see Merton, 1957, In recent years, a number of studies have investigated 1968) empirically the factors that have a significant effect on cita- tion counts (Bornmann & Daniel, 2008; Bornmann, Schier, The social constructivist view of citing is grounded in Marx, & Daniel, 2012). Montpetit, Blais, and Foucault the constructivist sociology of science (e.g., Collins, 2004; 2040 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—October 2012 DOI: 10.1002/asi
Knorr-Cetina, 1981; Latour & Woolgar, 1979). This view citation theories both refer to a limited range of factors. The casts doubt on the assumptions of normative theory and normative theory stresses factors that relate to the quality of questions the validity of evaluative citation analysis. Con- a work, such as the intellectual content of a paper, and the structivists have argued that the cognitive content of articles social constructivist view of citing emphasizes factors that has little influence on how the articles are received. Scien- have nothing to do with academic publishing conventions tific knowledge is socially constructed through the manipu- (e.g., citations as a way to make the content convincing to lation of political and financial resources and the use of the reader). Depending on the theory used, a successful rhetorical devices (Knorr-Cetina, 1991). For this reason, paper that has achieved a high citation impact is attributed citations cannot be satisfactorily described unidimensionally with specific characteristics that have led to its success. through the intellectual content of the article itself. The With the AKP, the two citation theories can be brought probability of being cited depends on factors that do not together in a broader context: According to the AKP, a paper have to do with the accepted conventions of scholarly pub- is successful and will be highly cited if a number of key lishing. Gilbert (1977), who has been particularly associated factors are fulfilled—one (important) factor being the with the social constructivist view, saw citations as an aid to quality of the paper. Another factor, for instance, is that persuasion, finding that scientists prefer to cite documents interest in a paper will be increased if at least one of the that are supportive of what they write and was preferably authors is well known. In addition, a short and succinct written by noted experts. paper with an appealing title is good for reception among Whereas Cronin (1984) found the existence of two com- colleagues. The extent to which these and other criteria peting theories of citing behavior hardly surprising, as the (both quality criteria and reputation/circulation criteria) are construction of scientific theory is generally characterized fulfilled should be reflected in the citation counts of scien- by ambivalence, for Liu (1997) and Weingart (2005), the tific publications. If one of these factors is lacking (e.g., a long-term oversimplification of thinking in terms of two high-quality paper is not published in a prestigious interna- theories reflected the absence of one satisfactory and tional journal), the citation frequency of the paper will most accepted theory on which better informed use of citation probably not be above average. indicators could be based. Liu (1997) and Nicolaisen (2003) In the area of citation counts, the AKP is related to saw the dynamic linkage of the two theories as a necessary what is called the principle of antidiagnostics, a medical step in the quest for a satisfactory citation theory. Four term that Braun and Schubert (1997) applied to the area of studies have empirically investigated the validity of the two bibliometrics: theoretical approaches. In agreement with the constructive view of citing behavior, Collins (1999) suggested that politi- While in medical diagnosis numerical laboratory results can cal and economic forces within the research process led to indicate only pathological status but not health, in scientomet- some papers being ignored by scientists whereas some were rics, numerical indicators can reliably suggest only eminence picked out. In contrast, studies by Baldi (1998), Stewart but never worthlessness. The level of citedness, for instance, may be affected by numerous factors other than inherent scien- (1983), and White (2004) provided more support for a nor- tific merits, but without such merits no statistically significant mative interpretation of the allocation of citations than for a eminence in citedness can be achieved. (p. 177) social constructivist interpretation. Implications of the study by Baldi have been discussed extensively in the literature in The meaning of this principle is that citation counts tend to recent years. Cronin (2004) assessed the Baldi study as “an better indicate a high quality of scientific papers than a low important and methodologically rigorous study” (p. 44). quality. A paper that has not received many citations is not Borgman and Furner (2002) found further comparison of automatically a paper of low quality. The reasons for the citing behavior within different disciplines necessary if we lack of citations can be many (discussed earlier). are to determine how far Baldi’s results may be generalized. Likewise, Small (1998) was not completely convinced by The AKP in Scientific Progress the results and stated: “A direct empirical test of the two theories seems difficult, and we need to take a step back and It is still not clear today how scientific progress functions view these two theories in a broader context” (p. 143). in detail and what the prerequisites for scientific break- That broader context could be the AKP as a combining throughs are. According to philosopher of science Thomas way of thinking about citation impact and behavior. The Kuhn, the development of science takes place in a cyclical formulation of a broader context follows approaches like pattern. Kuhn (1962) made a distinction between “normal those of Cole (1992) that serve the purpose of a rapproche- science,” which is oriented toward recognized explanatory ment between Merton’s (1973) sociological theory of models and methods (paradigms), and scientific revolutions, science and constructivist sociology of science. Although which become necessary when puzzling deviations can no most empirical studies examining factors influencing cita- longer be solved by existing paradigms. These anomalies tions (discussed earlier) have shown that there are always first lead to model drift and then to model crisis, when several factors that have an important impact (e.g., number “extraordinary science” begins. Here, the existing para- of authors of a publication, quality of a work, and reputation digms become blurred and are then replaced by new para- of the journal in which the paper is published), the two digms (paradigm change). These are the milestones of JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—October 2012 2041 DOI: 10.1002/asi
research, or scientific revolutions. They start up a new phase credit, and for what? . . . . These questions are far from simple. of normal science, but without cumulatively approximating It is commonly accepted among historians and philosophers of closer and closer any absolute truth. To illustrate, Kuhn science that most discoveries cannot be neatly localized in (1962) used the metaphor of biological evolution, which is space and time. They are not individual events, but complex and often messy processes extended over a period of time and similarly driven by problem solutions and does not advance involving many actors. Many scientific discoveries consist of toward any fixed goal. According to Kuhn (1962), the old several, more or less connected insights that in the end result in and new paradigms are incommensurable—that is, there is a consensus as to how the discovery has been made. A discovery no mutual understanding between the old and new ways of does not necessarily require a discoverer or a discovery event. looking at things and between the old and new terminology. (p. 142) Although several natural scientists have criticized the incommensurability thesis (e.g., Weinberg, 1998), the importance of Kuhn’s (1962) theory has never been funda- Here, Kragh and Smith (2003) pointed to something that mentally called into question (Hoyningen-Huene, 1993). we would like to clearly define in this article. We will Whereas Kuhn (1962) described that there is a turn of the attempt in the following to name some prerequisites for tide from normal science to scientific revolution, he did not scientific revolutions, adhering closely to the complex describe what the prerequisites must be for a revolution to process of the development of science. It is not our intention take place: to formulate the precise set of prerequisites but to introduce the AKP (see the Introduction). We would like to bring in the Rather we must explain why science—our surest example of AKP as a way of thinking about scientific breakthroughs: sound knowledge—progresses as it does, and we must first find Via the step-by-step fulfillment of several prerequisites, the out how, in fact it does progress. Surprisingly little is yet known emergence of scientific breakthroughs can be explained in a about the answer to that descriptive question. A vast amount of better way. In other words, the principle does not contain thoughtful empirical investigation is still required. (Kuhn, 1977, any previously overlooked new conditions for scientific p. 289) revolutions but instead delivers a certain view of that which leads to a breakthrough. The starting point of scientific activ- But by naming important criteria for assessing the quality of ity that develops into a scientific revolution is usually efforts new theories, Kuhn (1977) already discussed what is basi- to understand fundamental connections and to answer the cally needed for an old theory to be replaced by a new associated core questions (e.g., how did the universe or the theory: surface of the earth come into being?). Chance discoveries often play an important role in these activities, in that they What, I ask to begin with, are the characteristics of a good trigger events or drive development forward (Merton & scientific theory? Among a number of quite usual answers I Barber, 2004). But the prerequisites for scientific revolutions select five, not because they are exhaustive, but because they are are always factors such as independently produced empirical individually important and collectively sufficiently varied data that lead to similar results, the formulation of sound to indicate what is at stake. . . . These five characteristics— hypotheses that can be tested, and the development of a accuracy, consistency, scope, simplicity, and fruitfulness—are plausible theory that others acknowledge to provide an all standard criteria for evaluating the adequacy of a theory. If they had not been, I would have devoted far more space to them explanation of a phenomenon. Only when these and further in my book, for I agree entirely with the traditional view that prerequisites are fulfilled is there a “happy family” (in this they play a vital role when scientists must choose between an case, a scientific revolution), which is like all other happy established theory and upstart competitor. Together with others families in that all prerequisites are fulfilled. of much the same sort, they provide the shared basis for theory In the following, with a view to basing the discussion of choice. (pp. 321–322) AKP on typical examples taken from modern science, we refer repeatedly to some of the most significant scientific However, these basic criteria for assessing theories are revolutions in the 20th century in the area of the core disci- too abstract to allow us to name the prerequisites for para- plines in the natural sciences: the development of modern digm change in a discipline and to more closely examine the cosmology (Example 1), the development of modern geo- process of scientific progress, especially under the complex physics in the form of plate tectonics (Example 2), and the conditions of modern research. beginnings of quantum physics (Example 3). The starting Kuhn (1962) fully explained the complexity of scientific points in these areas of research were some of the core revolutions, taking the example of the Copernican Revolu- questions in the natural sciences: How did the cosmos tion. Kragh and Smith (2003) underlined the complexity develop? What is matter made of? How do the chemical and also in connection with modern cosmology—the discovery physical interactions of matter work? What is the chemical of the expanding universe: basis of metabolism and the heredity of living things? What forces and events shaped the surface of the earth? Is it at all reasonable to ask who made this discovery, and when? Before we come to some prerequisites that should be The expansion of the universe became recognized as a fact, but fulfilled for scientific revolutions, the three examples will be can it be described as a discovery? If it can, who deserves the introduced in brief. 2042 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—October 2012 DOI: 10.1002/asi
Example 1 direction. In the early 1960s, Hess (1962), and indepen- dently also Dietz (1961), developed a new theory of mantle Modern cosmology began at the start of the 20th century dynamics, which was based on thermal convection currents with Albert Einstein’s (1905b, 1915, 1916, 1917) special and in viscous mantle rock. general theories of relativity, which revolutionized our The rock magnetism data and the proposed mechanism understanding of space and time and introduced a new theory were brought together in the Vine–Matthews (1963) hypoth- of gravitation. In the 1920s, mathematical considerations led esis on the mobility of the earth’s crust. Wilson (1963, 1965) to the finding that the paradigm of a static universe can be produced further indications, independent of the magnetic replaced by the paradigm of a dynamic universe (Friedmann, data, and considerably expanded the theory of the moving 1922, 1924; Lemaitre, 1927). Before this, initial evidence surface of the earth. Wilson (1965) was the first to name the already had been found that called the notion of an eternally masses of moving with convergent boundaries “plates,” and static universe into question: Slipher (1912, 1917) found the he divided the earth’s outer layer into at least six large and redshift of the stellar spectral lines, which was the starting several smaller moving plates, which are kept in motion by point for the discovery of the receding galaxies by Hubble convection currents in the earth’s mantle. Seismology (the (1929) and Hubble and Humason (1931). Hubble and science of earthquakes) helped the plate tectonics model Humason combined a new method of cosmological distance achieve the final breakthrough, in that it showed that the measurement with Slipher’s data and also with their own boundaries of the postulated plates are identical to known measurements of the redshift and found a linear relationship earthquake zones. For the first time, there was now a com- between distance and velocity of the galaxies. plete picture regarding the events and forces that shaped the The cosmic microwave background radiation discovered surface of the earth. With plate tectonics, a new field of by Penzias and Wilson (1965) was immediately interpreted as research arose that dealt with large-scale motions of the a relic of an event when matter as well as space and time came earth’s crust (see Marx & Bornmann, in press). into being. In connection with the discovery of the redshift, this finally led to the paradigm of the Big Bang and an expanding universe. The origin and the frequency distribu- Example 3 tion of the chemical elements and the distribution of radio Quantum physics began with the discovery in 1900 of the galaxies fit the Big Bang model very well. However, it took a quantum of action (Wirkungsquantum) by Max Planck. The long time to verify an important prediction of the model: The quantum of action is an important physical constant; it says uneven distribution of matter in the universe in the form of a that physical quantities, such as energy, can take on only strong concentration of matter in isolated galaxies demanded discrete values. Max Planck (1900a, 1900b) discovered it, a slight irregularity already at the beginning of the evolution but restricted quantization to the exchange of energy of the cosmos. These beginning fluctuations in the density of between matter and electromagnetic field. Planck saw quan- matter must have been imprinted on the cosmic microwave tization as a temporarily necessary mathematical trick to background radiation and thus be provable as a pattern of allow theoretical treatment of the spectrum of black-body small temperature differences. When measurements by sat- radiation (the color of hot objects such as electric light ellite revealed just this pattern in the 1990s, the Big Bang bulbs). Einstein (1905a) applied the concept of quantization model was finally confirmed and became today’s recognized to light itself and, with his light quantum hypothesis, he cosmological standard model (see Marx & Bornmann, 2010). convincingly explained the photoelectric effect, which is the emission of electrons from matter as a consequence of their absorption of energy from light. For that explanation, Ein- Example 2 stein received the Nobel Prize in Physics in 1921. Based on Modern geophysics began at the beginning of the 20th Einstein’s hypothesis of light quanta, quantum theory century with Alfred Wegener. Examining the question of the became the basis of quantum mechanics in the mid-1920s. origin of the continents, Wegener found that there is a close Quantum mechanics describes the behavior of energy and fit between large-scale geological features of the coastlines matter at the atomic and subatomic scales (i.e., atomic and of separated continents. The matching features were an indi- nuclear physics) and became one of the main pillars of cation that the continents may have been joined together at modern physics. Quantum electrodynamics (QED) devel- one time. Because Wegener (1912, 1915) could not present oped out of quantum mechanics in the 1940s. Today, QED is any plausible mechanism for his theory of “continental the accepted paradigm of electromagnetic radiation and its drift,” it at first found little support among professional interaction with charged particles. colleagues. Further evidence for continental drift was found starting only in the mid 1950s, with paleomagnetic work on Prerequisites the direction of magnetization of rocks (Mason, 1958; Mason & Raff, 1961; Runcorn, 1955). This direction was Next, with the aid of these three examples, we present some apparently caused by the movement (spreading) of the sea of the conditions that we assume must be fulfilled for a scien- floor in connection with periodic polar reversals of the tific revolution to occur in a field. As prerequisites, solid earth’s magnetic field, which gave the rock its magnetic evidence in answer to basic questions must be presented (a), JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—October 2012 2043 DOI: 10.1002/asi
which is taken up by colleagues (b) and can be verified by papers (Marx & Bornmann, 2010). Following publication of means of independent data and methods (c). Especially impor- Wegener’s (1915) book on continental drift, reactions came tant is the statement of verifiable predictions that can then be quickly, even if they were at first overwhelmingly critical confirmed (d). This requires techniques for gathering data (e). (Example 2). The pioneering works of Planck (1900a, Of decisive importance are satisfactory and convincingly for- 1900b) and Einstein (1905a) were the start of a discussion mulated paradigms that answer more questions than they raise still ongoing today on the physical laws in the microcosm (f). This demands simplicity, elegance (aesthetic qualities) (g), and their philosophical aspects (Example 3). Outstanding and explanatory power of the new paradigm (h). For under- discoveries usually are soon taken up by colleagues, and standing of the paradigm, clear and plain language and the they often create excitement, creating an atmosphere of introduction of easy-to-remember labels are very helpful (i). departure (or a scientific “gold rush”). Finally, what is needed is the last crucial step that leads to the But that is not always the case. Sometimes new evidence definitive breakthrough and the establishment of a new para- or ideas are at first noted and not followed or pursued digm (j). For a scientist making a crucial contribution to a further. Whether new ideas are taken up depends essentially scientific revolution, we find stubbornness in thinking as well on whether they are put forward in an already existing dis- as good networking among colleagues in the field (k) to be cussion (e.g., the core questions in a field) or whether they important. The importance of these prerequisites for the occur- are difficult for colleagues in the field to interpret. In a rence of scientific revolutions will be illustrated in the follow- discipline, reserved judgment on ideas that can lead to a new ing using the three examples outlined earlier. paradigm is not only due to the fact that science must nec- essarily be conservative so it does not have to constantly turn Prerequisite 1: Solid evidence. Efforts to answer core to new paradigms. Thus, it is not always blindness or obsti- questions in science are not only driven forward by not- nacy that sometimes makes researchers react cautiously (or understood facts and connections but also steered in a new not at all), but rather practicality. direction via solid evidence (empirical evidence, or initial For instance, Einstein’s (1905b) new interpretation of evidence not yet confirmed independently). Submitting this space and time and his new theory of gravitation (1915, evidence and interpreting it in view of a possible new para- 1916, 1917) at first overtaxed many colleagues in the field digm to explain a phenomenon open up the matter to scien- (Example 1). The discovery of the quantum of action by tific dispute. The initial evidence that could speak for a new Planck (1900a) also was at first very disconcerting to col- paradigm is usually still quite vague and not abundant. The leagues (and to Planck himself) (Example 3). The discovery postulated receding of the galaxies (Example 1) was at first of the redshift by Slipher (1912, 1917) (Example 1) and the based exclusively on observation of the redshift in the emis- magnetic lineations or stripes on the ocean floor by Mason sion spectra of stars in the galaxies; for a long time, only (1958) and Mason and Raff (1961) (Example 2) are further empirical proof of the continental drift hypothesis (Example good examples of overtaxing: At the time of the discoveries, 2) was the measurement of rock magnetization. The techni- these researchers and their colleagues did not know quite cal options for gathering further evidence are often still what to make of them. But the discoveries found their way underdeveloped, which is why independent confirmation at into the scientific archives (the professional journals and first fails to appear. Initial evidence that speaks for a new databases) and could be used later for the research. paradigm is frequently questioned because, for the most part, there is a lot of room for interpreting it within the Prerequisite 3: Verifying evidence from independent framework of the old paradigm. For instance, the discovery research groups. Indications and initial evidence are of the redshift (Example 1) also could be explained within usually based on still-sparse data. Further evidence found by the paradigm of a static universe, as de Sitter (1917) pro- independent research groups is an important prerequisite for posed. This model was based on a different geometry than a field to develop a new paradigm. In Kuhn’s (1977) later that of Einstein’s (1915, 1916, 1917) also-static universe, work, he commented on how this evidence helps in the and it seemed to allow for redshift without requiring a new choice between competing theories: paradigm of an expanding universe. Analogously, the direc- tion of magnetization of rocks could have been changed by Measurement can be an immensely powerful weapon in the battle between two theories, and that, I think, is its second chemical processes (Example 2), so that it had little to do particularly significant function. Furthermore, it is for this with the magnetic field of the earth at the time the rock was function—aid in the choice between theories—and for it alone, formed (and thus little to do with rock moving). that we must reserve the word “confirmation.” (p. 211) Prerequisite 2: Interest among colleagues who take up on For Kuhn (1977), new empirical data and continuous theory the ideas. When Hubble (1929) and Hubble and Humason development are essential for progress in a field: (1931) published the linear relation between distance and velocity of galaxies, discussion was opened among col- One recurrent implication of the preceding discussion is leagues in the field (and beyond) on the paradigm of a that much quantitative research, both empirical and theore- dynamic universe (Example 1). This is shown, among other tical, is normally prerequisite to fruitful quantification of a things, by the relatively high number of citations of these given research field. In the absence of such prior work, the 2044 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—October 2012 DOI: 10.1002/asi
methodological directive, “Go ye forth and measure,” may the islands by radiometric dating completely confirmed well prove only an invitation to waste time. (p. 213) Wilson’s assumptions. An expedition of the drilling ship Glomar Challenger systematically and comprehensively Not uncommonly, new evidence pertinent to answering measured the age profile of the sea floor over the Mid- fundamental questions has been independently produced Atlantic Ridge. The result: The age of the sea floor, as by different researchers, at the same time (or also at dif- expected, increases symmetrically with increasing distance ferent times) and in separate lines of work (sometimes from the ridge to the east and west. This was the clinching even in different, unrelated subdisciplines). For instance, at evidence that new magma from deep within the earth rises the start of the development of plate tectonics (Example 2), at the structurally weak joins between the plates that make both land- and ocean-based magnetic measurements were up the earth’s crust and eventually erupts along the crest of made. Later, the different lines of research were combined. the ridges to create new, spreading oceanic crust, transport- This happens frequently if a researcher having a broad, ing the continents along with it. interdisciplinary overview recognizes the lines of work and uses them as the basis for a new paradigm. Of course, the Prerequisite 5: Suitable techniques for the required mea- prerequisite for this is that a number of pieces of evidence, surements. Sometimes there are no techniques or technolo- produced in different subdisciplines based on different gies available for exact empirical testing of hypotheses that experimental methods, point in the same direction. could lead to a new paradigm. Measurement of irregular With data from independent research groups, there is a cosmic microwave background radiation, which would much higher probability that a new paradigm based on concur with an expanding universe, became possible only initial evidence is in fact valid. However, the independent with space travel (Example 1). The demonstration of inho- evidence must be recognized as the elements that can mogeneity in the microwave background radiation was then confirm a theory. This is not always the case immediately. a convincing confirmation of the Big Bang model. Without In geophysics (Example 2), the sea floor magnetic stripes the existence of these fluctuations, it would not be possible had not been looked for in a targeted manner at all after to explain the unequal distribution of matter in the universe, the discovery of the remanent magnetization of rocks on the birth of galaxies, and thus our own existence. The devel- land (Creer, Irving, & Runcorn, 1957; Runcorn, 1955), but opments in plate tectonics were very closely connected with were instead found completely independently of that dis- the technical advances of wartime (World War II) and the covery and rather by chance (Mason, 1958; Mason & Raff, postwar period (Example 2). Probes such as the echo 1961). The connection between these observations and sounder and magnetometer were decisive prerequisites for suppositions concerning crustal mobility were recognized large-scale measurements of the structure and magnetization only later. However, sometimes in efforts to verify a (new) of the sea floor. Besides the mechanism responsible for paradigm, there is a deliberate search for evidence that can continental drift, the result of the drift had to be investigated: confirm or contradict the paradigm. For instance, the The predicted slow drift of the continents could only be spatial irregularity of cosmic microwave background radia- measured precisely, and then finally proven, only later tion (the insignificant fluctuations in temperature that the through the use of satellites. theory of a dynamic universe requires) (Example 1) was not found by chance but instead systematically in the Prerequisite 6: A theory that provides a plausible explana- framework of an expensive satellite program. tion of the empirical findings. In the interplay of theory and experiment that essentially shapes scientific advance- Prerequisite 4: The paradigm should make possible ment, what is always needed in addition to evidence in the (correct) predictions. The usefulness of a paradigm is form of empirical data is a plausible interpretation of that measured in terms of the extent to which it can yield correct data in a (new) theory. The theory must not only describe predictions. There is a good example of this in cosmology but also explain how something works (e.g., how the earth’s (Example 1): Cosmic microwave background radiation was continents move). Thus, the theory must contain the exact found by chance by Penzias and Wilson (1965); however, it mechanism in the form of driving powers and processes. had been predicted many years previously, and Alpher Regarding the hypothesis of continental drift (Example 2), (1948) and Alpher and Herman (1949) had even calculated Wegener’s approach (1912, 1915) was not recognized by its wavelength. In this connection, the prediction of fluctua- colleagues in the field because he had no convincing mecha- tions in the background radiation is no less impressive than nism for how continents move. He thought that tidal forces the successful empirical observation. Geophysics (Example caused by the moon might be responsible. However, that 2) also provides an example here: The Canadian geophysi- was not possible quantitatively. A convincing reason for cist Wilson (1963) supposed that the age of the islands of continental drift was presented only a half-century later by Hawaii increases with their distance from the East Pacific Hess (1962): material heated by radioactive elements in the Rise. An essentially stationary hotspot of erupting magma earth’s interior and the resulting convection currents. As had apparently created a trail of volcanic islands, which Hess’s proposal that the sea floor itself moved, or spread, then drifted with the moving crust, and their active volcan- was unorthodox at the time, he cautiously called his theory ism gradually stopped. Determination of the age of rock on “geopoetry.” JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—October 2012 2045 DOI: 10.1002/asi
With the overwhelming evidence and many premises, it Prerequisite 9: The paradigm has a catchy name. Appeal- was only a question of time before an unbiased (young) ing names for a new paradigm (e.g., Big Bang or plate tecton- researcher (Fred Vine) brought together Hess’s (1962) ics) make it easier for the scientist (and persons outside of theory and the already-available evidence (Mason, 1958; science) to name things and put things in a nutshell. Authors Mason & Raff, 1961) in a convincing paradigm: the Vine– that introduce attractive names in the specialist literature do Matthews hypothesis (Vine & Matthews, 1963). The the discipline a good turn. A catchy name is important because hypothesis brought together the idea of the sea floor spread- when a new paradigm is being established, there is not yet any ing and the observed phenomenon of the magnetic striping generally accepted, uniform language usage. In the debate patterns. This interplay of theory and empirical data con- over the Big Bang model versus the competing Steady State nected previously separate lines of work, thus preparing for model (Example 1), the term “Big Bang” was coined, rather the later development of plate tectonics. Many other disparagingly, by an opponent of the theory (Fred Hoyle) researchers also working in these areas had not been as during a talk on a BBC radio program in 1950. The catchy versed in both geology and physics as were Vine and Mat- phrase caught on, in both camps, and it got right to the heart of thews and had been viewing the connections from too the new paradigm. When Wilson (1965) published a paper on narrow a perspective. The Vine–Matthews hypothesis was the transform faults (plate boundaries, at which the plates an important prerequisite for further progress in geophysics, move in relation to one another) (Example 2), he named the as it was a plausible and solid theory of the supposed, very moving rock masses “plates.” Wilson divided the earth’s slow movement of the earth’s crust. surface into several major and minor plates that are kept in movement by convection currents in the earth’s crust. In the article, Wilson (1965) gave the new paradigm an attractive Prerequisite 7: The paradigm is simple and elegant. The name, calling it “plate tectonics.” persuasive power of a new paradigm has a lot to do with whether it is simple and elegant (Kuipers, 2002; McAllis- Prerequisite 10: The last crucial step is achieved. In con- ter, 1996). Invited to write a self-chosen inscription on a nection with the history of modern cosmology (Example 1), blackboard after holding a lecture at the University of the question arises as to why Hubble (1929) and Hubble and Moscow in 1956, Paul Dirac wrote, “a physical law must Humason (1931) are generally acknowledged to be the dis- possess mathematical beauty” (Kragh, 1990, p. 275). Dirac coverers of the receding of the galaxies and not, for instance, (1928a, 1928b) used this premise when he derived the Slipher (1912, 1917), who much earlier had found decisive existence of the antielectron (positron, the first antimatter initial evidence. Although they did not venture to interpret particle) from his wave equation for the electron (the Dirac their findings as demonstrating an expanding universe, equation) and his hole theory based on it. The antielectron Hubble (1929) and Hubble and Humason (1931) had was discovered shortly thereafter by C.D. Anderson (1932) achieved the crucial last step and presented their findings in the cosmic radiation (Example 3). The Big Bang model clearly and unmistakably in words and images (in contrast to (Example 1) provided an elegant explanation of the origin their forerunners): and structure of the visible universe. The paradigm of plate tectonics (Example 2) is very satisfactory aesthetically, as But surely it is true that Hubble’s measurements, added to the it provides simple and compelling answers to most of the data of Slipher (and Humason) provided the crucial last step to basic questions in the geosciences. finally put together the first full recession graph. In other words, Hubble’s work may have formed only part of the contribution, but it was the crucial last brick—a common occurrence in Prerequisite 8: The paradigm has great explanatory science! (C. O’Raifeartaigh, personal communication, October power. Prior to the 1960s, there was no generally accepted 17, 2010) paradigm for the evolution of the surface of the earth (Example 2). There was no satisfactory answer to the ques- In analogous fashion, publication of the Vine–Matthews tion of the origin of continents, oceans, mountains, valleys, hypothesis (Vine & Matthews, 1963) can be seen as the and volcanoes. The answers came only starting in the 1960s crucial last step for establishing the paradigm of plate tec- with the paradigm of plate tectonics, which found general tonics (Example 2). The articles by Hubble (1929), Hubble acceptance around 1970. The paradigm came about as the and Humason (1931), and Vine and Matthews (1963) have result of bringing together various subdisciplines of geo- a strong synthesis character, as they bring together previ- physics. The many branches of science that make up the ously separate lines of research: In Hubble’s case, the mea- study of the earth (geomagnetism, seismology, petrography, surement of cosmic distance and measurement of the and geophysics) remained at first fragments. It was the new spectral shift of galaxies, and in Vine and Matthews’s case, paradigm of plate tectonics that brought them together (A.H. the striped pattern of natural remanent magnetism and Anderson, 1971). The explanatory power of the Big Bang measurement of the heat flows that can be explained by model (Example 1) results from the many points of contact mantle convection. with classical astronomy and high-energy physics. The sci- entific community recognized the great explanatory power Prerequisite 11: The researcher has stubbornness in think- of the Big Bang very early on. ing and good networking with colleagues in the field. As a 2046 JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—October 2012 DOI: 10.1002/asi
You can also read