Friday 4 January 2008

UK scientists down-shift to second rate research

Charlton BG, Andras P. ‘Down-shifting’ among top UK scientists? - The decline of ‘revolutionary science’ and the rise of ‘normal science’ in the UK compared with the USA. Medical Hypotheses. 2008; 70: 465-472 - doi: 10.1016/j.mehy.2007.12.004

[I apologize that the tables are not properly formatted - but I believe they can be understood with some extra effort. The Figure is missing from this pre-print, but it merely summarizes Table 1].

Editorial

‘Down-shifting’ among top UK scientists – The decline of ‘revolutionary science’ and the rise of ‘normal science’ in the UK compared with the USA

Bruce G. Charltona, , Editor-in-Chief – Medical Hypotheses and Peter Andrasa
aMedical Hypotheses, Newcastle University, Henry Wellcome Building, NE1 7RU, UK

Available online 28 January 2008.



Summary
It is sometimes asserted that UK science is thriving, at other times that it has declined. We suggest that both assertions are partly true because the UK is thriving with respect to the volume of ‘normal’ science production but at the same time declining in the highest level of ‘revolutionary’ science. Revolutionary science may be distinguished from normal science in that revolutionary science aims at generating qualitative advances which change the direction of established science, while ‘normal’ science aims at incremental progress extrapolating from established science. Revolutionary science has been measured by counting national numbers of science Nobel laureates and ISI Highly Cited (HiCi) scientists; normal science has been measured using the total volume of scientific publications and citations at both national and institutional levels. By these criteria the UK has been progressively catching-up with the USA in terms of normal science since the 1990s. At the same time the UK has declined in revolutionary science over recent decades by a significant brain drain of future Nobel laureates and HiCi scientists, and a sharply reduced success (both in absolute and compared with the USA) at winning science Nobel prizes. One possible cause for this pattern could be a time-lag, such that the UK’s improved science production since about 1990 may eventually work-through into improved UK performance in revolutionary science. More pessimistically, this pattern may reflect a strategic down-shift of the best UK-resident scientists away from revolutionary science and towards less-ambitious and safer normal science which is more productive in the short term.

Introduction

It is sometimes asserted that UK science is thriving, at other times that it has declined. We suggest that both assertions are partly true, and that the UK is thriving with respect to the volume of science production and at the same time UK performance has declined at the highest level of ‘revolutionary’ science.

We suggest that the UK pattern of catching-up with the US in normal science while declining in revolutionary science can be explained by reduced ambition among the best UK scientists – a ‘down-shift’ from aiming at a breakthrough in revolutionary science and towards incremental progress in normal science.

Revolutionary science versus normal science

Revolutionary science is a term coined by Thomas Kuhn in his book The Structure of Scientific Revolutions (Chicago University Press, 1970) to describe research which changes the fundamental structures of science by making new theories, discoveries or technologies (i.e., new ‘paradigms’).

Revolutionary science may be a breakthrough which changes the fundamental structures of a whole science (as achieved by Einstein, Newton or Darwin) or, more often, which develops a significant sub-speciality of a major science (as Crick and Watson developed molecular biology). Revolutionary science is therefore the cutting-edge which solves problems that are intractable to the incremental gradualism of normal science, and thereby allows each science to continue to grow in rapid bursts, and over the long term to become qualitatively more accurate and useful in its predictions.

But normal science does not attempt to establish new directions or to develop qualitatively new explanations or technologies, or to make paradigm-shattering discoveries. Instead, normal science is the incremental extrapolation of already-existing paradigms – and it comprises building on established research by procedures such as checking, trial-and-error and gradual improvement in the precision of measurement.

The great bulk of research in modern developed societies such as the USA, UK, France, Germany, China or Japan is ‘normal science’ [1]. Normal science attempts to improve on established science and to extend its capabilities step-by-step further in the direction towards which it is already tending. This means that normal science can be planned and managed. Normal science is also amenable to evaluation by peer-review processes because much science is so modest in its ambition, and its methods are so well-tested, that even newly-published research can be regarded as ‘pre-validated’, and may be ready for implementation without the necessity for further corroboration [2].

The intrinsically high risk of failure in revolutionary science
Revolutionary science aims at generating qualitative advances – new theories, techniques or discoveries which change the direction of established science, while by contrast normal science aims at incremental progress extrapolating from established science.

Since revolutionary science is much more ambitious than normal science, it is intrinsically riskier and more likely to fail in its aims. The biographies of even the most prestigious revolutionary scientists typically contain many failures both before and after the success for which they became famous – for example Crick and Watson’s botched DNA model of 1951 [3], and Andrew Wiles’ mistaken announcement of 1993 that he had proved Fermat’s last theorem [4]. In both cases, despite these psychologically-devastating setbacks, the scientists continued to work on the problem and produced a correct solution within a couple of years. But the fact that even the very best scientists fail when doing the most ambitious work emphasizes the riskiness of revolutionary science.

Since outcomes are so unpredictable, revolutionary science should be defined by its aspirations rather than by its outcomes. A system of revolutionary science therefore consists of the scientists actively working on a problem of potentially-revolutionary scope and the scientific communications between these scientists (including all their relevant verbal and written communications: conversations, e-mails, letters, formal publications, specialist journals, seminars, conferences, etc.). Recognized revolutionary scientists should therefore be regarded as merely the clearest and most visible tip of a much larger iceberg of ambitious researchers who are engaged in endeavours to solve tough scientific problems. So although the majority of revolutionary scientists never succeed in their objectives and never attain fame or success, their co-operation, competition, checking and critique nonetheless plays a crucial role in the process of scientific advance [5].

By contrast with revolutionary science, normal science is much more likely to succeed than to fail, especially when normal science is conducted by able and well-trained scientists working in large multi-disciplinary teams located in well-resourced institutions. In this sense, normal science resembles industrial research and development, which can reliably accumulate numerous small incremental improvements built upon already-established science.

Of course there is an overlap between revolutionary and normal science, for example when one of the increments of progress in normal science unexpectedly turns out to have revolutionary implications. This is ‘serendipity’, but even here chance favours only the prepared mind, and scientific happy accidents may be neither noticed nor exploited unless discoverers recognize their potentially revolutionary implications. On the whole, cautious, modestly-ambitious research which aims at incremental extension of existing knowledge along established lines is very unlikely to lead to a scientific revolution, and if something hinting at a scientific revolution does crop–up it may well be ignored because it was unpredicted, and not part of the plan.

In a nutshell, the greater short term productivity and greater chance of a successful outcome that are characteristics of normal science are attainable only at the cost of diminution in the potential importance of successful research [2].

Measuring revolutionary and normal science

Science production can be measured and analyzed quantitatively using standard scientometric research outputs such as number and share of publications and citations counted over a period of time; with credit allocated to a specified unit of production such as a nation, an institution, a research group or an individual scientist. This activity comprises the discipline of scientometrics [6], [7] and [8].

But successful revolutionary science is rare – indeed entirely absent from many research situations – and in modern, developed societies revolutionary science is swamped by the much larger volume of normal science [5] and [8]. Furthermore, successful revolutionary science can usually only be detected retrospectively and after a time lag of many years. In the short term it is not possible accurately to pick-out the successful revolutionary science from the much larger pool of speculative theories, apparently anomalous observations and radical suggestions for technological improvements.

Only after the revolutionary science has proven itself in terms of triggering new developments which have spread through into normal science can it be confidently assumed that a scientific revolution has indeed occurred. For example, the occurrence of revolutionary science can be retrospectively detected by observing a new direction or change in the direction of research; which may be visible in terms of new types of subject matter for study, new methods or technologies, new conference topics and journals, and eventually new types of research unit and educational specializations within universities (such as new modules or degree programmes).

The problem of discriminating between revolutionary and normal science has become even more difficult since the advent of Big Science [1] and [6]. Big Science comprises quasi-industrial forms of research organization. It arose initially in physics during the 1939–1945 world war (for example the development of radar in the UK or the much larger Manhattan project for developing the atomic bomb in the USA); but Big Science organization now characterizes biomedical research [9], which is currently the dominant world science.

Big Science is almost inevitably a type of normal science, since it needs to be planned and predicted, hence it must be modestly incremental. Furthermore, Big Science tends to be ‘applied’ in its aims, and similar to industrial Research and Development in its methods.

Therefore, different scientometric methods are needed to detect and measure the rare but potentially more-important examples of paradigm-transforming Kuhnian ‘revolutionary science’. We suggest that scientometric measurement of revolutionary science might focus on identifying and counting successful revolutionary scientists instead of measuring the total of scientific production.

Normal science trends – UK versus US

Normal science can be measured using the total number of scientific publications or citations to publication listed in databases such as Thomson Scientific’s Web of Knowledge (http://portal.isiknowledge.com) or Elsevier’s Scopus (www.scopus.com). While the total number of citations and publications includes revolutionary science as well as normal science, it seems likely that – given the rarity of successful revolutionary science – the proportion of publications and citations attributable to revolutionary science from the total of all publications and citations will not be insignificant.

Using data on number of scientific publications and their citations it seems that the UK (and also Europe generally) has been increasing its market share relative to the USA over recent years. For example, King reports [10] that from 1993–1997 to 1997–2001 the percentage UK share of total world publications was second only to the USA (the UK has since been overtaken by China [11]), and increased from 9.29 up to 9.43 while citations increased from 10.87 up to 11.39. At the same time, US percentage shares were declining for both publications (37.46 down to 34.86) and citations (52.3 down to 49.43). So this represents a significant catch-up of the UK with the US in normal science production throughout recent years.

To examine normal science at an institutional level, we measured science production of publications and citations at the most successful UK and US scientific research universities using the ISI Web of Science (WoS) Science Citation Index (SciCitI) data for the period 1975–2004. We determined publication and citation counts for 94 UK universities and 299 US universities (national universities and top-50 liberal arts colleges – derived from US News) for all years between 1975–2004.

The WoS was searched for each university and each year to determine the number of publications published by members of the given university in the given year and the number of citations that these papers received since they were published up to the time of data collection (February 2006). Total counts for five year periods were pooled to reduce yearly fluctuations. This method means that each publication and its citations will be multiply-credited to all institutions that are listed in its author affiliations.

UK and US universities were ranked for the six five-year periods for publication and citation counts. The top-20 universities were determined for each of these 12 UK rankings. For the top-20 UK universities in each UK ranking we calculated their average rank in the corresponding UK–US ranking (see Table 1 and Fig. 1). This generated a statistic for the average UK–US rank of the top-20 UK universities at each five year period for SciCitI-listed publications and citations.

Table 1.

Average ranking (to nearest integer) of total volume of publications and citations for top-20 ranked UK universities when included with 299 top-ranked US universities and colleges 1975–2004 in five year segments 1975–1979 1980–1984 1985–1989 1990–1994 1995–1999 2000–2004
Publications average rank UK top-20 81 81 79 77 65 62
Citations average rank UK top-20 76 83 84 83 70 65


Figure 1. Average ranking of top-20 UK universities relative to 299 top US universities in five year segments 1975–2005.


The table and graph demonstrate that the top US universities are much more productive of publications and citations than are the top UK universities, since the average top-20 UK university never ranks higher than 62nd place at any timepoint. Therefore, on average the best US universities are much more productive of normal science than the best UK universities.

But again the UK seems to be catching-up with the US in recent years. The average top-20 UK University showed improving average UK–US ranking in both publications and citations from the 1975–1979 period to the 2000–2004 period. Thirty years ago the average top-20 university would be ranked in 81st place for publications and 76th place for citations whereas most recently this average top-20 UK university has moved up 19 places in the rankings for publications and 11 places in the ranking for citations.

On closer inspection, the UK citation ranking worsened in the first 15 years, then sharply improved in 1990–2004 period with an average top-20 UK university improving by 19 places relative to the US (see Fig. 1 and Table 1).

The conclusion from both national-level and institutional-level Web of Science data is that in terms of normal science production the UK is worse than the USA, but the UK has been catching-up over the past two decades.

Revolutionary science trends – UK versus US Nobel prizes

The scientometrics of revolutionary science is vestigial; and the subject has until recently been the domain of historians of science who have worked by detailed and discipline-specific study of documentary material, sometimes supplemented by retrospective interviews [5], [12] and [13].

But this historical method suffers from being labour intensive and piecemeal; and simpler, more quantitative methods would be desirable. Here we have measured revolutionary science by counting successful individual revolutionary scientists using science Nobel prizes and data on the migration of ‘Highly Cited’ (HiCi) scientists from the Thomson Scientific database (previously the Institute of Scientific Information – ISI).

The award of a Nobel prize in one of the four recognized sciences (Physics, Chemistry, Physiology/Medicine and Economics) seems likely to be the best current evidence of a significant achievement in revolutionary science [14]. We are making the assumption that national success at generating Nobel-quality revolutionary science in these four scientific domains is indicative of success across a broader range of top-notch science. Although the small annual number of Nobel prize-winners (‘laureates’ – maximum of three per discipline equal to twelve per year) means that many significant achievements go unrecognized; nonetheless the perceived validity of these awards is high within the scientific community, and only a small proportion of awards are regarded as controversial or unjustified.

However, it must be remembered that counting Nobel laureates only measures the most highly-visible and most fully-validated tip of a presumed iceberg of revolutionary science. The prize credits successful revolutionary science which has changed the direction of a discipline in a big way, and where credit for this can be allocated to a single person or a few individuals. It is almost certain, on general theoretical grounds derived from complex systems theory [15] and [16], that the process of generating major breakthroughs in revolutionary science must be supported by a much larger submerged base of revolutionary science research which is harder to identify with confidence, and where credit for achievements is spread between many individual scientists.

When the number of Nobel laureates in the UK and the US are counted for 20 year segments [17] we can see that the UK has suffered a sharp decline in the past twenty years, from 25 prizes 1967–1986 to just 9 prizes from 1987 to 2006. By contrast the US has increased the number of laureates from 88 in 1967–1986 to 126 in 1987–2006. In other words, over a period of 60 years the UK has declined from being broadly equal to the USA in terms of Nobel prizes per capita, and holding undisputed second place (by a large margin) in terms of total Nobel prizes, down to the current position of winning similar numbers of Nobel prizes to other comparable large developed nations such as Germany (9 prizes) or France (5 prizes).

To illustrate this, we can observe that in the past 20 years, the USA has 16 institutions which have won three or more prizes, compared with a complete absence of such treble-Nobel institutions in the UK. From 1967 to 1986 University of Cambridge, the MRC Molecular Biology Unit at Cambridge, University of Oxford and Imperial College, London won an impressive total of 17 Nobel prizes between them – however since 1986 they have all-together won just 3 [17].

In a separate study of the official biographies and autobiographies of Nobel laureates on http://nobelrpize.org, Charlton examined the pattern of national migration for the past 60 years (where this could be established). Each laureate’s prize was allocated to a nation on the basis of the working address at the time the prize was awarded. Some laureates were omitted because their biographies were incomplete, some laureates were retired, and some worked at unclassifiable multinational units such as CERN.

Table 2 shows the numbers of US versus UK laureates for 20 year segments 1947–2006. ‘Total immigrant laureates’ is the number of scientists who graduated from university elsewhere then migrated either to the US or UK where they were working at the time when they received the Nobel prize. This data suggests that in the past 20 years the UK has lost its previous ability to attract future Nobel-prize-winning scientists from elsewhere.

Table 2.

US:UK science Nobel laureates – immigration and migration trends 1947–2006 1947–1966 1967–1986 1987–2006
Total laureates included 45:20 85:24 112:9
Total immigrant laureates 11:3 25:6 21:0
UK–US migration 0 5 5



The row ‘UK–US migration’ shows the number of scientists during each 20 year segment who were educated in the UK then migrated to the USA where they were awarded a Nobel prize. (The reverse US to UK migration did not happen during the past 60 years.) This trend suggests that the US has become increasingly attractive to UK educated scientists over the past 60 years, such that for 1987–2006 five out of fourteen (36%) of all UK-educated laureates had moved to the USA by the time they won the Nobel prize.

UK versus the US – Highly Cited scientists
Nobel laureates might be regarded as a very partial and biased sample of successful revolutionary scientists; although Nobel differentials are maintained when other comparably prestigious prizes, medals and awards are added to the analysis (e.g., for disciplines such as clinical medicine, mathematics and computing science [18] and [19]). A more objective and complete sample of high status scientists can be derived from scientometric data concerning the most frequently-cited academics in specific scientific disciplines.

Highly Cited (HiCi) scientists are defined by Thomson Scientific as those researchers who have received the highest number of citations in their field of research [20]. Highly Cited status is likely to be less correlated with revolutionary science than winning a Nobel prize, since a scientist can accumulate many citations without making a revolutionary breakthrough but instead by exceptionally high productivity, or by doing research that is very useful while not being revolutionary in import. Nonetheless, many HiCi academics do go-on to win Nobel prizes and similar awards [7].

Ioannidis reported that during 1981–1999, 56% of UK-born HiCi academics had migrated to another country compared with only 2% of US born HiCi academics [21]. This demonstrates that the US is much more able to retain its best quality scientists than is the UK. In amplification of these results, Ali et al. have recently examined migration and productivity of HiCi academics in economics, physics and biosciences – which confirms that the US is a net gainer of HiCi academics while the UK is a net loser [22]. Furthermore, a preliminary study by Pierson and Cotgreave, which examined a cohort of UK science Ph.D.’s from 1988 to 2000, suggested that those who had migrated to the USA were able to generate more citations per article than those who remained in the UK [23].

In conclusion, studies of the migration patterns of Highly Cited scientists in the Thomson Scientific (previously ISI) Web of Science database are consistent with the Nobel laureate data in demonstrating that the US is nowadays a much more significant focus of revolutionary science than the UK. While the field remains undeveloped, the pattern of scientometric evidence which plausibly measures national attainment in revolutionary science strongly indicates that the UK has substantially declined relative to the USA.

In other words, although the UK remains probably the third most productive scientific country in the world, the UK has now declined in revolutionary science to the point where the nation is unable either to attract or to hold-onto the very best revolutionary scientists – whether these are measured by HiCi status or by Nobel prizes.

Time-lag or down-shift

Our conclusion is that the UK has been progressively catching-up with the USA in terms of normal science production since about the 1990s, but the UK has sharply declined in revolutionary science achievements over recent decades. Optimistic and pessimistic interpretations of this pattern are possible.

An optimistic interpretation would be that the pattern is the result of a time-lag between improvement in normal science and revolutionary science. In this case the divergence between normal and revolutionary science would be regarded as temporary, and simply a consequence of insufficient time for the short term improvement in normal science since about 1990 to work through into improved performance in the longer-term indices measuring revolutionary science. If this optimistic interpretation is correct, the UK should soon see a rapid increase in the number of Nobel laureates and the Highly Cited scientists, a reversal of the brain drain of the best UK scientists, and also a renewed ability of the UK to attract the very best scientists from around the world.

However, on balance we would favour a more pessimistic interpretation of these results since the above analysis indicates a significant brain drain of the very best UK scientists [21] and [22]. It is likely that this loss of talent has contributed to a reduced UK level of activity in revolutionary science research. But the magnitude of decline in UK revolutionary science may indicate that additional factors are at work, such that the top quality scientists who remain in the UK may have redirected their efforts from revolutionary science to normal science.

One hypothesis is that there may have been a ‘down-shift’ strategy of the most-able UK scientists to direct their efforts into solving easier and less-important scientific problems than they are capable of tackling. For reasons discussed earlier, the numbers of ‘most-able’ scientists is likely to be large and comprise easily enough people to influence scientometric measures (for instance, we have informally observed that the science production of a single large research team is sufficient to move a UK university up by several places in the rankings for publications and especially citations – unpublished observations).

Probably, there are many hundreds of potentially-elite professional scientists working in the UK. Their choices, such as the choice between working on revolutionary or normal science problems, will have a large impact precisely because their productivity is so much higher than average. For instance, we suggest that potential UK Nobel prize-winners may, over recent decades, have re-orientated their research away from the riskier strategy of pursuing revolutionary science and towards less ambitious projects that are more immediately productive. In other words we propose that these top UK scientists may – on average - have shifted down a gear, to accelerate their careers by solving more, smaller or easier problems over the short term.

Since these are some of the ablest and best-trained scientists in the world, by down-shifting into normal science they are typically able to out-perform their competitors in terms of producing large quantities of high quality normal science. This would have the effects of enhancing scientometric measures of normal science at the top universities specifically and also boosting national level statistics for publications and citations. This move improves the prospects of a swift accumulation of grant income to invest in manpower and equipment, and fuel further research that would tend to attract citations early in the scientist’s career.

A strategy of top scientists increasingly eschewing risk and pursuing the more certain rewards of normal science could therefore account for the UK pattern of declining revolutionary science and improving normal science.

Conclusion

Our conclusion is that UK revolutionary science has declined over the past sixty years. Yet at the same time the production of UK science as a whole has actually grown faster than US science in terms of the number of publications and citations. Our speculative interpretation of this pattern of simultaneous rise and fall focuses upon the down-shifting of ambition of UK scientists. We also assume that a similar down-shift has not occurred to the same extent in the USA.

Given that revolutionary science is a high risk and long-term endeavour which usually fails, it is likely to thrive only when the incentives rewarding success are much greater than for normal science. Our hypothesis entails that the incentives to encourage the best scientists to pursue revolutionary science would be more powerful in the USA [24]. Such US incentives might in principle include less severe punishment for failure when aiming high in science; and greater rewards for success in revolutionary science than for success in normal science. Such rewards for revolutionary scientists might include higher pay, more favourable working conditions, greater chance of employment at elite institutions, longer-term grant support, and a bigger chance of winning prestigious prizes.

To test this hypothesis requires empirical investigation. For example, matched cohorts of the best UK and US scientists could be interviewed and compared in terms of their career choices, scientific ambitions and their perceived incentives. We predict that the best US scientists would demonstrate a stronger orientation towards working in revolutionary science than would the best UK scientists. The position, pay and conditions of successful revolutionary scientists over recent decades could be compared with the most successful normal scientists. We predict that in the USA the rewards for revolutionary science were relatively greater compared with normal science than in the UK.

Is the decline of the UK in revolutionary science a genuine cause for concern We believe it may be. The whole scientific world benefits from US achievements in revolutionary science, but the sheer scale of US dominance in revolutionary science may contain the seeds of its own destruction. Since the short term incentives will always favour normal science, there seems to be a potential danger of lack of international competition eventually leading to declining US standards of revolutionary science in the long term.

We suspect that over recent decades the UK has become an increasingly-efficient factory for producing normal science at a high quality and volume. But apparently the UK no longer specializes in revolutionary science in the way that it did until recent decades; and the UK now mainly serves as an incubator of talented personnel who must usually transfer to the US to fulfil their scientific potential. To regenerate the UK as a base for revolutionary science would probably require increasing the incentives that would reward success in revolutionary science, thereby encourage greater scientific ambition and risk-taking.

Acknowledgement

Thanks are due to Neil Herald for his work on the analysis of the top-20 UK university production relative to the US.

References

[1] J. Ziman, Real science, Cambridge University Press, Cambridge, (UK) (2000).

[2] B.G. Charlton, Conflicts of interest in medical science: peer usage, peer review and ‘CoI consultancy’, Med Hypotheses 63 (2004), pp. 181–186 [Editorial].

[3] J.D. Watson In: Stent Gunther, Editor, The double helix, Weidenfeld & Nicolson, London (1981).

[4] S. Singh, Fermat’s last theorem, Fourth Estate, London (2002).

[5] D.L. Hull, Science as a process, Chicago University Press, Chicago (1988).

[6] D.J. de Solla Price, Little science Big Science: and beyond, Columbia University Press, NY (1986).

[7] Garfield E. Do Nobel prizewinners write citation classics Current Contents 1986;23:3–8. . Accessed 30 Nov 2007.

[8] B.G. Charlton and P. Andras, Evaluating universities using simple scientometric research-output metrics: total citation counts per university for a retrospective seven-year rolling sample, Science and Public Policy 34 (2007), pp. 555–563.

[9] B.G. Charlton and P. Andras, Medical research funding may have over-expanded and be due for collapse, QJM 98 (2005), pp. 53–55.

[10] King DA, The scientific impact of nations, Nature 430 (2004), pp. 311–316.

[11] Leydesdorff L, Wagner C. Is the United States losing ground in science Scientometrics, in press. Accessed 30 Nov 2007.

[12] H.F. Judson, The eighth day of creation: makers of the revolution in biology, Jonathan Cape, London (1979).

[13] D. Healy, The antidepressant era, Harvard University Press, Cambridge, (MA, USA) (1998).

[14] B.G. Charlton, Why there should be more science Nobel prizes and laureates, Med Hypotheses 68 (2007), pp. 471–473.

[15] N. Luhmann, Social systems, Harvard University Press, Cambridge, (MA, USA) (1995).

[16] B. Charlton and P. Andras, The modernization imperative, Imprint Academic, Exeter, UK (2003).

[17] Charlton BG. Scientometric identification of elite ‘revolutionary science’ research institutions by analysis of trends in Nobel prizes 1947–2006. Med Hypotheses 2007;68:931–4.

[18] Charlton BG. Which are the best nations and institutions for revolutionary science 1987–2006 Analysis using a combined metric of Nobel prizes, Field medals, Lasker awards and Turing awards (NFLT metric). Med Hypotheses 2006;68:1191–4.

[19] Charlton BG. Measuring revolutionary biomedical science 1992–2006 using Nobel prizes, Lasker (clinical medicine) awards and Gairdner awards (NLG metric). Med Hypotheses 2007;69:1–5.

[20] Thomson Scientific. ISIHighlyCited.com. Accessed 30 Nov 2007.

[21] J.P.A. Ioannidis, Global estimates of high-level brain drain and deficit, FASEB J 18 (2004), pp. 936–939.

[22] Ali S, Carden G, Culling B, Hunter R, Oswald AJ, Owen N, Ralsmark H, Snodgrass N. Elite scientists and the global brain drain. Paper presented at the World Universities Conference, Shanghai, China Octerber 2007. Accessed 30 Nov 2007.

[23] A.S. Pierson and P. Cotgreave, Citation figures suggest that the UK brain drain is a genuine problem, Nature 407 (2000), p. 13.

[24] B.G. Charlton and P. Andras, The future of ‘pure’ medical science: the need for a new specialist professional research system, Med Hypotheses 65 (2005), pp. 419–425.

Tuesday 1 January 2008

Bronowski's principle of tolerance

Editorial

Charlton BG. Jacob Bronowski’s principle of tolerance. Medical Hypotheses.
2008; 70: 215-17.

Summary

In The principle of tolerance, Jacob Bronowski discusses a vital but neglected characteristic of science: that ‘‘all information is imperfect’’, and ‘‘our ability to work and act in the real world depends on our accepting a tolerance in our recognition and in our language’’. The nineteenth century ideal that “science should speak the perfect factual truth has turned out to be inaccessible”. But this should not be a cause for regret, because “if things had to be identical before you could recognize them, you would never recognize anything at all”. The principle of tolerance is the judgment that two instances are sufficiently similar that we can treat them as the same for present purposes. “Tolerance – is the essential safeguard, the essential degree of coarseness which makes it possible to work with abstract entities in the real world”. Too much tolerance and you are misled by random variation; too little tolerance and you lose valuable information. The most beneficial degree of tolerance must be a matter of judgment because it cannot be determined in advance. So, the best level of tolerance is known only retrospectively, by comparing the rate of progress of science when a greater or lesser degree of tolerance is assumed. The judgment of tolerance which led to the fastest scientific progress is justified as having been the best. Science therefore needs to tolerate different judgments of tolerance among scientists, allowing a multiplicity of levels of tolerance to coexist and compete. Bronowski’s principle of tolerance locates the roots of science in the domain of human creativity, in the necessity for personal judgment in science, and in the provisional and progressive nature of scientific truth: “You have to tell the truth the way you see it. And yet you have to be tolerant of the fact that neither you nor the man you are arguing with is going to get it right”.

***

In The principle of tolerance, Jacob Bronowski (1908–1974) discusses a vital but neglected characteristic of science: that ‘‘all information is imperfect’’, and ‘‘our ability to work and act in the real world depends on our accepting a tolerance in our recognition and in our language’’ [1].

The nineteenth century ideal that “science should speak the perfect factual truth has turned out to be inaccessible”. But this should not be a cause for regret, because “if things had to be identical before you could recognize them, you would never recognize anything at all”; and “if we were given the superhuman power to identify things only when they are identical, it would be fatal for us”. Indeed, the scientist “…would not be able to do any experiment at all. [He] would keep on saying to a colleague, “You are not doing it right. It is not the same experiment”.

Bronowski’s point is that two experiments are never exactly the same – and if we insisted on exactness nothing could ever be replicated. The principle of tolerance is the judgment that two instances are sufficiently similar that we can treat them as the same for present purposes. “…Tolerance – is the essential safeguard, the essential degree of coarseness which makes it possible to work with abstract entities in the real world”.

I recognize this phenomenon from my first days as an active scientist, training for the doctorate. I was learning how to perform radio-immunoassays (RIAs) to measure peptides – which was a standard methodology. My initial reaction was shock at how imprecise, how subjective, was this supposedly ‘standard’ method. There were many personal judgments required to generate each measurement, and each judgment involved a trade-off.

The usual practice was to measure each plasma sample in duplicate and average the result. I felt this was not sufficiently precise, but the more replicates used the more plasma was used; which in turn meant fewer blood tests were possible for each experimental subject – or else each subject would need to give a bigger blood sample, reducing the pool of subjects.

Counting the radioactivity of each plasma sample took a long time, and the lower was the concentration of peptide, the lower was the level of radioactivity and the larger the stochastic variation of decay; but the longer the counting procedure, the fewer experiments could be done. Defining the sensitivity limit of the assay was another judgment call. If I set the sensitivity too low I would just be measuring random noise, but if I set the threshold too high I would be missing data on differing levels of peptides.

And this is the problem in microcosm. Too much tolerance and science becomes un-reliable [2] because you are misled by random variation; too little tolerance and you lose valuable information. The too-credulous scientist may be spinning a story based on foundations of sand; but the ultra-sceptic will block progress by knocking-down every potential advance. And different scientists will make different judgments concerning the optimal level of tolerance. As so often, it is clear that a variety of personality types are necessary in the social process of science [3] and [4].

Naïve observers of science tend to regard science as characterized by the elimination of judgment and the attainment of absolute precision, but Bronowski points out that: “[The scientific process] depends on an understanding that the best scientific result in the world is not right, that the best experiment in the world is surrounded by an area of tolerance [1]”.

The key point is that the optimal degree of tolerance must be a matter of judgment because it cannot be determined in advance. The best level of tolerance is known only retrospectively, in practice, by comparing the rate of progress of science when a greater or lesser degree of tolerance is assumed. Levels of tolerance therefore compete. The judgments of tolerance which (further down the line) lead to the fastest scientific progress are justified after the fact as having been the best judgments [4]. This also suggests that science needs to be tolerant of different degrees of tolerance among scientists themselves [3], and allow a multiplicity of levels of tolerance to coexist and compete.

There are thus two pressures in science. One is to strive for ever-lower levels of tolerance and to seek ever-greater levels of precision in measurement. The other is to accept a reasonable, attainable level of tolerance so that work can proceed now.

Too much tolerance and science will be measuring random noise, too little tolerance and the progress of science will be stalled. On the one hand, science tries to be correct and accurate; on the other hand, all science is wrong, in an ultimate sense, because it will be superseded – and the proper question is whether current scientific practice is accurate enough. We need to compromise over tolerance in order to act.

Over the long term science tends to progress by becoming more precise, by including more data, and with a diminishing area of tolerance. For example, an improved technology may increase the precision of a measurement; that – indeed – was the purpose of my doctorate, to deploy a new and better immunoradiometric assay (IRMA) for the hormone ACTH. Such new and more precise measurements may contradict predictions, meaning that existing theories are challenged. A new, more complex and more-inclusive theory may be devised.

This long term trend towards lower tolerance tends to award greater short-term status to those sceptics of science who are biased in the direction of reducing tolerance, discarding data, rejecting theories. Scientists who fear being wrong will tend to err in the direction of being in-tolerant. The tendency is to generate a supply of professional nay-sayers and negativists among the community of scientists.

Another way to avoid personal criticism for mistaken judgments is to remove decisions out of the realm of judgment and submit them to convention. Whenever an interpretational decision is removed from judgment to convention it is also removed from science, and ceases to be a part of scientific communications. This tendency to standardize and routinize is an intrinsic aspect of the evolution of complex systems [5]. Once they are standardized, procedures need to be learned but do not need to be thought-about or justified: this liberates time, resources and effort to devote to developing the cutting-edge of new science. In other words, conventions of tolerance are not science but should be the result of science, and may increase the efficiency of science.

[Science] “is a process in which truth and falsity (as well as other principles, values, goods or evils) are there not in the observation itself but in the record of it which you pass to others [1]”.

In the end, therefore, subjective evaluations of tolerance are those which are intrinsic to scientific communications, and even the decision to adhere to conventional practice is itself a judgment.

However, balancing the short-term tendency to accord status to low-tolerance, nay-saying ultra-sceptics; is a longer-term tendency of science to award the greatest prestige to those yea-sayers who open up new growth areas of revolutionary science. Because, if tolerance becomes too low, this will tend to kill-off a science. When practice becomes too difficult and new information too rare, that branch of science will dry-up through lack of anything to communicate.

So in the long-term the yea-sayers ultimately overcome the nay sayers of science, from the intrinsic tendency of complex systems to perpetuate themselves [5]. Tolerance levels tend to relax sufficiently to enable scientific communication to be generated and sustained, and to grow.

Bronowski’s principle of tolerance locates the roots of science in the domain of human creativity, in the necessity for personal judgment in science, and in the provisional and progressive nature of scientific truth:

“…You have to tell the truth the way you see it. And yet you have to be tolerant of the fact that neither you nor the man you are arguing with is going to get it right [1].”

References

[1] J. Bronowski, The principle of tolerance, A sense of the future, MIT Press, Cambridge, MA, USA (1977).

[2] J. Ziman, Reliable knowledge, Cambridge University Press, Cambridge, UK (1978).

[3] B.G. Charlton, From nutty professor to buddy love: personality types in modern science, Medical Hypotheses 68 (2007), pp. 243–244. SummaryPlus | Full Text + Links | PDF (68 K) | View Record in Scopus | Cited By in Scopus (0)

[4] D.L. Hull, Science as a process, Chicago University Press, Chicago (1988).

[5] B. Charlton and P. Andras, The modernization imperative, Imprint Academic, Exeter (2003).