Where Is Science Going?
Diana M. Hicks and J. Sylvan Katz
Science Policy ResearchUnit, Universityof Sussex
Do researchersproducescientificand technicalknowledgedifferentlythan they did ten
years ago? Whatwill scientific researchlook like ten years from now? Addressingsuch
questions means looking at science from a dynamic systems perspective. Two recent
books about the social systemof science, by Zimanand by Gibbons,Limoges,Nowotny,
Schwartzjnan,Scott, and Trow, accept this challenge and argue that the research
enterprise is changing. This article uses bibliometricdata to examine the extent and
natureof changes identifiedby these authors,takingas an exampleBritishresearch.We
use theirtheoreticalframeworksto investigatefive characteristicsof researchsaid to be
increasingly pervasive-namely, application, interdisciplinarity,networking,internationalization, and concentrationof resources. Results indicate that research may be
becoming more interdisciplinaryand that research is increasinglyconductedmore in
networks,both domesticand international;but the data are moreambiguousregarding
applicationand concentration.
Introduction
In PrometheusBound: Science in a Dynamic Steady State, John Ziman
examines science facing a future"withina fixed or slowly growingenvelope
of resources" (Ziman 1994, 10). For 300 years, science expanded very
quickly-at an exponentialrate.Exponentialgrowthcame to be seen as the
normby scientists,but its continuancewas absurd;eventually,it would have
led to every man, woman,andchild spendingall theirtime writingscientific
papers(Ziman 1994, 9-10). As the demandsof science on resourcesbeganto
conflict with otherpriorities,provisionof resourcesslowed and approached
a steady state; painful adjustmentswere needed in a system that evolved
AUTHORS'NOTE:We would like to thankNigel Ling for researchassistance.We also would
like to thank the following for financial support:U.K. Office of Science and Technology,
Departmentof Tradeand Industry,Departmentof Health,MedicalResearchCouncil,Engineering and Physical Sciences ResearchCouncil, and the Economic and Social ResearchCouncil.
We are grateful to MargaretSharp,Ben Martin,RikardStankiewicz, Keith Pavitt, and Slavo
Radosevic for helpful discussions and to threereferees for theircomments.
Science, Technology,& HumanValues,Vol. 21 No. 4, Fall 1996 379-406
? 1996 Sage PublicationsInc.
379
380
Science, Technology,& HumanValues
underconditionsof exponentialgrowth.Ziman tracesthe consequences for
researchers,primarilyacademic,and for the advanceof knowledge.
MichaelGibbons,CamilleLimoges,HelgaNowotny,SimonSchwartzman,
Peter Scott, and MartinTrow,authorsof The New Productionof Knowledge
(1994), arguethattheinternaldynamicsof sciencehavegenerateda new way of
producingknowledge.The post-WorldWarII expansionof the researchand
educationsystems,coupledwith the inexorablelogic of entrepreneurial
fundraising,broughtintobeinga methodof performingresearchqualitativelydifferent fromthe discipline-basedacademicactivitythathas dominatedscience and
ourthinkingaboutscience.They call this new formof research"mode2."
The two analyses synthesize disparatestrandsrunningthroughcomment
on the research enterprise, describing the interplay among many factors.
They conclude that internaldynamics are bringing about a transitionto a
different type of science system, and they focus our attention on science
as an evolving system of interacting institutions. In so doing, they
suggest that examining systemic-level data about science could be important,andthey providea theoreticalframeworkto help interpretsuch data.
In this article, we use system-wide bibliometricdata to examine several
characteristicsof the changing research system described by Ziman and
Gibbonset al.
Ideally, systemic data about science would be comprehensive,covering
all research institutionsin a nation, and would include the links between
institutions, enabling us to track their evolution over time. Bibliometric
analysiscan generatesuch databecausethe Science CitationIndex(SCI),the
basis for most bibliometricanalyses,approachescomprehensivecoverage of
high-quality,international,publishedresearchoutput;it containscitationand
coauthorshiplinks; and it goes back some forty years. But does the scientometric literaturecontain systemic analyses as defined above? Not really.
Bibliometric analyses usually examine nations as a whole or output from
particularprogramsor in certainfields or froma smallgroupof institutionsthe top twenty,for example.In largepart,this is because analystshave been
limited by the techniques available for handling large text databases at a
reasonableprice. However,technicaladvanceshave permittedus to attempt
such an analysison a medium-sizedscientificcountry-the UnitedKingdom,
with 8 percentof the world'soutput.Ouranalysisof the U.K. science system
is a longitudinal,comprehensive,institutional-levelprofile of U.K. research
output.With these data,we are able to addresssome aspectsof the evolution
of sciencesystemsdiscussedby ZimanandGibbonset al.
Hicks and Katz / Whereis Science Going?
381
Hypotheses
Both books arguethatthe researchenterpriseis undergoinga transition,
analyzing many facets of an integratedsystem to supportthis thesis. Of
necessity, our analysis approachesthe materialslightly differently.We can
look for changes in the science system empirically,but we can do this only
by examiningin turneach facet of the overall argument.Therefore,we first
isolate the characteristicsof the science system the authorsbelieve to be
changing and list these as "hypotheses."These lists are not meantto reproduce the essence of the authors'theses, as the lists are a simplificationand
reduction of their arguments;rather,the lists are meant to act as a bridge
between their analyses and our data.
In PrometheusBound,Zimanarguesthatmodernscience is characterized
by the following:
1.
2.
3.
4.
5.
6.
7.
8.
9.
more management
more evaluation
careerstructureswith less permanence
sophisticatedinstrumentation
more emphasison application
greaterinterdisciplinarity
more networkingand collaboration
more internationalization
more specializationand concentrationof resources
We cannot examine all these points using bibliometricdata, so we will
focus on the last five: emphasizingapplication,interdisciplinarity,networking and collaboration,internationalization,and concentrationof resources.
Ziman believes these featureshave increasinglycharacterizedscience over
the last fifteen to twenty years.
They are also some of the features said to distinguish mode 2 research
from its predecessor.Gibbonset al. arguethatin the new mode of knowledge
production,
1. qualitycontrolis no longer simply a matterof peer review;
2. more knowledge is producedin the context of application;
3. consensus on solutions to scientific problemsis formed outside disciplinary
boundaries,generatingtransdisciplinaryknowledge;
4. research is performedby intellectually and organizationallyheterogeneous
groupsoften broughttogetheronly for the durationof the project;and
5. researchoutcomes areinfluencedby social accountability,andresearchersare
sensitive to the broaderimplicationsof theirwork, becoming more reflexive.
382
Science, Technology,& HumanValues
Again, we cannotexamineall of these hypothesesusing bibliometricdata;
instead, we will focus on three of them-context of application,transdisciplinarity,andthe natureof the groupsthatperformresearch.Underthe fourth
heading,the authorsput forwardthreemore hypotheses:
1. An increasein the numberof potentialsites whereknowledge can be created;
no longer only universitiesandcolleges, but nonuniversityinstitutes,research
centres, government agencies, industrial laboratories,think-tanks,consultancies, in theirinteraction.
2. The linkingtogetherof sites in a varietyof ways-electronically, organizationally, socially, informally-through functioningnetworksof communication.
3. The simultaneousdifferentiation,at these sites, of fields andareasof studyinto
finer and finer specialities. The recombinationand reconfigurationof these
subfields form the bases for new forms of useful knowledge. Over time,
knowledge productionmoves increasinglyaway from traditionaldisciplinary
activity into new societal contexts. (Gibbonset al. 1994, 6)
These hypothesesoverlapwith those of Ziman, and we believe they can
be addressed using systemic-level bibliometric data of the sort we have
developed. We can investigate research in the context of application by
comparinggrowthin publishingin more appliedfields with growthin more
basic fields, and by comparinggrowthin publishingby institutionsthatboth
produce and apply research with growth in publishing by those that just
produce research.We can judge the extent to which publishing might be
moving beyonddisciplinaryboundariesby examiningthe growthof publishing in cross-disciplinaryjournals. We can uncover the variety among publishing institutionsby counting the numberof institutionsof various types
that publish. We can assess whetherpublicationis becoming more or less
concentratedby investigatingthe distributionof articlesacross institutions.
And finally, our datapermitus to glimpse domestic and internationalscientific networks using as an indicator collaborative articles. We take each
hypothesis in turn, examining the bibliometricevidence from the United
Kingdomover the 1980s to assess the magnitudeof any changes.
Methodological Background
The bibliometricmethod used adheredto de facto standardsin the bibliometric community(Katz et al. 1995; Katz and Hicks 1995a). The difference between this and previous work is that every U.K. address on the
376,226 articles we processed was assigned to one of approximately5,000
unified institutionalnames. Thus, for every article producedin the United
Hicks and Katz/ Whereis Science Going?
Table 1.
383
List of Sectors and Notes on Their Definitions
Sector Name
Notes on Definitions
University
"Old"universities, polytechnics were a separate sector in
the 1980s, see below
No hospitals were counted as part of universities
Intramurallaboratories, excluding "groups"at universities
but including "units"at universities
Including all laboratories privatized during the decade
Special Health Authorityand British postgraduate medical
research institutesa
Departmental laboratories and local government
Does not include research funded by grants from charities,
in universities, for example
Sector became universities in the 1990s
Comprising other educational, other medical, and unknown,
each of which produces less than 2 percent of U.K. output
Hospital
Research council
Industry
SHA and BPG
Government
Nonprofit
Polytechnics
Other
a. These are special research hospitals, and so the hospital sector is effectively split
into two parts, for reasons of policy relevance.
Kingdom during the 1980s and indexed in the SCI, we know with which
institutions its authors were affiliated. Each of the institutions was assigned
to a sector, and the analysis presented here is at the sectoral level.
The data were derived from the SCI. Information on all published works
indexed in the SCI and listing a U.K. address from 1981 to 1991 was bought
from the Institute for Scientific Information (ISI) on tape. From the tapes,
three document types were extracted-articles, notes, and reviews (collectively referred to as "articles" herein), as these tend to report original research
results. Each publication was then processed to unify the institutional addresses and assign each institution to a sector. The following sectors were used:
university, hospital, research council, industry, Special Health Authority
(SHA) and British postgraduate medical institutes (BPG), government, nonprofit, polytechnics, and other. The definition of each sector is explained in
Table 1.
Application
The first hypothesis about the changing nature of research is that more
research is being performed within the context of "a need to be application
relevant." We propose to examine this hypothesis in two ways. First, we will
384
Science, Technology,& HumanValues
compare the growth in publicationby institutionsthat apply research and
those thatdo not. Second, we will comparethe growthof fields closer to and
fartherfrom application.
To performthe institutionalcomparison,we have identifiedindustryand
hospitalsas the two sectorsin which researchis bothperformedand applied.
If researchincreasinglyis performedin the context of application,we might
expect thatthe numberof articleslisting eithera companyorhospitaladdress
would be growing faster than the number of articles listing addresses of
institutionsin othersectors.'Table2 reportsthe growthin each sector's share
of publications.The table displays the following for each sector: the total
numberof articlespublishedfrom 1981 to 1991, the shareof the U.K. total
this represents,the averageyearly growthin share(which is the slope of the
linear regressionthroughthe yearly share data), and the ratio between the
yearly growthandthe sector's averageshareof U.K. articlesover the decade
(multipliedby 100). Since the sectors differin size quite a bit, the last figure
is presentedto normalizethe yearly growthrate for the size of each sector.
The sectors are sortedby this ratio,with the largestat the top of the list.
Table2 indicatesthatpublishingby hospitalsgrew the fastest, with their
shareof U.K. outputincreasingby 0.43 percentper year.2When adjustedfor
the size of sector, publishing by nonprofit institutions, SHA and BPG
institutesandhospitalsgrew the fastest.Publishingby industrygrew respectably,placingit in the middleof the list. Growthratesin the othersectorswere
lower. To the extent thatresearchperformedin the context of applicationis
indicated by articles listing a hospital or company address, the share of
publicationsproducedin the context of applicationis clearlyincreasing,and
this stands in contrastto the shareof publicationsproducedby universities,
polytechnics,researchcouncils, government,andothersmalleducationaland
medicalpublishers.Suchresearchdoes not appearto be increasingthe fastest,
however. The data suggest that biomedical publishing is increasing more
stronglythanpublishingconnecteddirectlywith application.
However, researchperformedby hospitals and industryis not the only
researchorientedtowardapplication.Therefore,we examine the same issue
by looking at the growthin publishingin fields thatarecloser to application.
We have classified articlesinto scientific fields basedon thejournalin which
they appeared.Our scheme is derived from that of ISI, which classified
journalsusingjournal-journalcitationpatterns,expertassessment,and feedback fromusers of specialtyproducts.We have aggregatedthe 154 subfields
of the ISI scheme into 11 fields in a mannersimilarthough not identicalto
that of ISI. However, ISI could not classify every journaluniquely into one
subfield; some journalshave more than one subfield assignment.In certain
cases, a journal'ssubfields will fall into differentfields. Insteadof fraction-
HicksandKatz/ Whereis ScienceGoing? 385
Table 2.
Publication Growth by Sector
Sector
Articles
Share
Nonprofit
SHA and BPG
Hospital
Industry
Other
University
Polytechnic
Research council
Government
8,878
17,448
81,719
28,088
9,832
217,441
8,008
42,814
15,597
2.4
4.6
21.7
7.5
2.6
57.8
2.1
11.4
4.1
Average Yearly
Growth in Share
0.05
0.10
0.43
0.08
0.01
0.12
-0.01
-0.10
-0.11
i 0.01
i?0.01
i 0.04
i 0.03
? 0.01
+ 0.06
i0.01
?10.05
?0.01
Yearly Growth/
Average Share
0
100)
2.3
2.2
2.0
1.1
0.3
0.2
-0.5
-0.9
-2.7
ating such journalsacross differentfields, we chose to createspecial "interfield" categoriesto containthem.3
Our classification scheme contains seventeen disciplinary and crossdisciplinaryfields (Katz and Hicks 1995a). The fields are furtherclassified
into four disciplinarygroups:life science, naturalscience, engineeringand
materials science, and interdisciplinary.Life science fields are medicine,
biology, agriculture,andinterfieldlife sciences (containingjournalsthatspan
two of the other life science fields). Naturalscience fields are chemistry,
physics, earthandspace sciences, mathematics,andinterfieldnaturalscience.
Engineeringand materialscience fields areengineering,materials,information andcommunicationtechnologies,andinterfieldengineeringandmaterials
science. Interdisciplinarycategories are multidisciplinary(containingenvironmentalsciences as well as Nature,Science, Proceedings of the National
Academy,etc.) andthreefields containingjournalsthatspantwo disciplines.
For the purposesof this analysis,we takefields closer to applicationto be
agriculture,medical, engineering,information,materials,and interfieldengineeringand materialsscience. Table3 reportsthe growthin publishingby
fields. As in Table2, the columns reportthe following: the total numberof
articles in each field, the share this representsof U.K. output, the yearly
change in share obtainedfrom the slope of a linear regressionthroughthe
yearly share data, and the ratio between the slope and the average share
(multipliedby 100). As above, the columns are sortedby this ratio.The table
indicates that medicine and materialscience are two of the fastest growing
subfields but also that agriculture,information,and engineeringaccounted
for a smallershareof U.K. outputat the beginningof the 1990s thanthey had
386
Science, Technology,& HumanValues
Table 3.
Growth in Publishing in Fields Closer to Application
Field
Interfieldnatural sciences
Material sciences
Natural sciences-engineering
and materials science
Medicine
Interfieldlife sciences
and
Life sciences-engineering
material science
Physics
Earth and space sciences
Interfieldengineering and
material science
Biology
Life sciences-natural sciences
Multidisciplinary
Engineering
Chemistry
Agriculture
Mathematics
Informationand communication
technologies
Articles
Share
(%)
Yearly Change/
Yearly Change Average Share
in Share (%)
(x 100)
2,590
5,503
1
1
0.07 + 0.02
0.04 + 0.01
10.5
2.4
10,063
140,012
26,931
3
37
7
0.04 + 0.02
0.49 i 0.06
0.09 + 0.03
1.6
1.3
1.2
1,081
33,702
9,105
0
9
2
0.00 i 0.01
0.04 + 0.05
-0.01 + 0.01
0.7
0.4
-0.5
4,788
46,533
7,985
16,545
9,919
38,474
13,635
6,918
1
12
2
4
3
10
4
2
-0.01 + 0.01
-0.11 +0.04
-0.02 + 0.01
-0.05 + 0.02
-0.04 + 0.02
-0.20 + 0.04
-0.13+0.02
-0.07 + 0.01
-0.6
-0.9
-1.1
-1.1
-1.7
-2.0
-3.7
-4.0
1,464
0
-0.05 + 0.01
-12.1
at the beginning of the 1980s. Thus the data on growth of fields does not
support the hypothesis that applied research is growing more quickly than
basic research. Overall, the results are somewhat ambiguous. The institutional analysis indicates that research in the context of application is increasing, in contrast with research produced by other sectors, but it is not increasing the fastest. The field analysis does not support a simple hypothesis that
more research is being performed in the context of application. Of course,
both indicators are crude. There are application-oriented research institutions
and groups in sectors other than hospital and industry, and there are applied
journals of chemistry and more basic journals of engineering.4 A finer grained
analysis might reduce some of the ambiguity. Nevertheless, this analysis does
warn against simplistic statements about the growth of applied research.
Hicks and Katz / Where is Science Going?
387
Transdisciplinary Publishing
The second hypothesis about the changing natureof the researchenterprise concernsinterdisciplinaryresearch.Many observersof science believe
thatsuch researchis increasinglypervasiveand significant.However,few if
any attemptshave been madeto measurethe relativeextent of interdisciplinary versus disciplinaryresearchactivity and the rates of change. This is not
surprisinggiven the apparentimpossibilityof even agreeingon a definition
of interdisciplinaryresearch.
Nevertheless, in the face of these difficulties, we have somewhatboldly
proposedto measurethe increasein interdisciplinarypublishingusing articles publishedin cross-disciplinaryjournalsas an indicator.We conjecture
that journals that cannot be classified into a single field or even a single
disciplinecross field boundariesin some way,perhapsby addressingmultiple
audiences,or by taking submissionsfrom several researchcommunities,or
by publishing articles that tend to requirea broaderrange of skills than is
customaryin a traditionaldiscipline.We label suchjournalsand the articles
We suggest thatthis transdisciplinary
publishedin them "transdisciplinary."
publishingbears some relationto the much discussed concept of interdisciplinaryresearch.
We will assess the growth in transdisciplinarypublishingby examining
whetherthe shareof articlespublishedin single-fieldjournals(those classified as medicine,biology, chemistry,or physics, for example)has decreased
and the share of articles in cross-field journals (those classified as, for
example, interfieldlife sciences, life sciences/naturalsciences, or multidisciplinary) increased. The second and third columns of Table 4 report the
numberof articlesand share of U.K. outputaccountedfor by each category
of journal.The fourthcolumn lists the yearly change in percentageshareof
U.K. outputcalculatedfromthe slope of theregressionline throughthe yearly
data.
The table indicatesthatthe shareof U.K. outputpublishedin journalsthat
were assignedto single fields such as medicineor physics decreasedslightly,
while the share of U.K. output published in journals that could not be
classified into single fields increasedslightly.The shareof articlesin singledisciplinejournalsdecreasedby 0.11 percentper year on average,while the
share in transdisciplinaryjournals increased by 0.11 percent per year.
Transdisciplinarypublishinghas severalcomponents.The shareof publishing that crosses field boundaries(for example, medicine and agriculture,
physics and chemistry)but remainswithinthe same discipline(life sciences,
naturalsciences, or engineeringand materialssciences) increasedby 0.14
388
Science, Technology,& HumanValues
Table 4.
Growth in Share of Interdisciplinary Articles
Category
Single-field journals
Cross-field journals
Number of
Articles
Share of U.K.
Output (%)
Yearly Change
in Share (%)
305,265
69,983
81.4 + 0.6
18.6 +0.6
-0.11 + 0.05
0.11 +0.05
Interfield
34,309
9.1+ 0.6
0.14 + 0.03
Interdisciplinary
Multidisciplinary
19,129
16,545
5.1 +0.3
4.4 + 0.2
0.02 +0.03
-0.05 + 0.02
375,248a
100
Total
a. Excludes articles in joumals classified as "unknown,"which represent only 0.3
percent of all U.K. articles.
percent per year, the largest increase. The share of articles in journals
that span disciplinary boundaries(i.e., life sciences/naturalsciences, life
sciences/engineering and material science, or naturalsciences/engineering and materials science) remained stable. Articles in multidisciplinary
journals such as Nature or Science are not themselves multidisciplinary.
Thesejournalscannotbe classified into a single discipline,however,because
they accept articlesfrom many disciplines. The shareof U.K. articlesin this
"multidisciplinary"category (which includes also environmentaljournals)
decreased slightly. The data support the hypothesis that publishing in
transdisciplinaryjournalsis playing a largerrole in U.K. science. However,
there has been no dramaticshift in the percentageof U.K. output in each
category.Roughly four-fifthsof U.K. outputis in single-disciplinejournals,
and roughly one-fifth is in transdisciplinary
journals.5
The data supportthe idea thatresearchcrossing the boundariesof traditional disciplines increasedmore quickly than disciplinaryresearchin the
United Kingdomduringthe 1980s. However,the differencesin growthrates
were not large enough to create a large increase in the share of U.K.
publicationsin cross-disciplinaryjournalsat the expense of publicationsin
traditionaljournals.Therefore,these dataindicatethatdisciplinaryresearch
still accountsfor by farthe bulk of U.K. scientific outputandwill apparently
continueto do so for some time to come. However,the measureis imperfect.
Certainfields, such as agricultureor materialsscience, are considered"disciplinary"in our scheme, but this designationcould be questioned.Refined
journalclassificationmightproducemoreintuitivelysatisfyingresults.Even
so, a journal-basedclassification will always be less satisfactorythan an
article-by-articleclassification,however impracticalsuch a scheme may be
to implement.6
HicksandKatz/ Whereis ScienceGoing? 389
Collaboration
Disciplinaryand interdisciplinaryresearchdiffer in partin the diversity
of skills andknowledgebroughtto bearon researchproblems.A widerrange
of capability is combined to produce an interdisciplinaryarticle than a
disciplinary article. A greater breadth of skills and knowledge may be
demandedby researchin the fastest growingareasof science-or so the last
section suggests. The extent to which skills and competenciesarecombined
to producescientific articlescan be assessed in otherways. In particular,we
can examine the extent to which authorsand institutionswork together to
produceand publish researchby examining collaborativearticles. Because
complementaryassets such as knowledge, equipment, and materials are
broughttogetherby collaboratingscientists, increasesin collaborationindicate more combining of knowledge and equipment to produce scientific
articles.
Both ZimanandGibbonset al. point to increasingnetworking,collaboration, and communicationin the scientific community,arguingthatnetworking is in partnecessitatedby the increasinginterdisciplinarityof researchand
in itself facilitates interdisciplinarity.As with interdisciplinarity,many observers of science have commented upon growth in collaboration. In
addition, analysts have produced bibliometricevidence of increasing research collaboration.De Solla Price (1986) noted increasingcollaborative
publishingas long ago as 1963. More recently,the bibliometriccommunity
has focused on increasesin internationalcollaboration(EuropeanCommission 1994; Leclerc et al. 1992; Luukkonen,Persson, and Sivertsen 1992;
Narin and Whitlow 1990). Our data permitthe first detailed assessment of
the rate of collaborationamong institutionswithina nationalscience system
and its growthover time.
We begin our examination of collaboration and the extent to which
institutionalresourcesare combined in researchby calculatingthe average
numberof authors,institutions,and countriesinvolved in the productionof
an article.We find thatby the end of the decade,an averagenumberof authors
for a U.K. articlewas 3.4, and the averagenumberof domestic institutions
was 1.3. The averagenumberof foreign countriesfor an articlethatlisted a
foreign address was 1.3. The increase in these averages is analyzed using
linearregression,andthe resultsaredisplayedin Table5. This tableindicates
that the extent to which resources were combined to produce scientific
articlesgraduallyincreasedduringthe 1980s.
We can examine these trendsin greaterdetail by looking at the frequency
distributionof authors,institutions,and countriesper article(see Figure 1).
The graphs display these distributionsin 1981 and 1991.7The first graph
390
Science, Technology,& HumanValues
Table 5.
Average Number of Authors, Institutions, and Countries
per Article
1981
Authors
Institutions
Countries
2.63
1.19
1.17
1991
Linear Yearly
Increase
3.34
1.28
1.25
0.08 + 0.01
0.01 + 0.00
0.01 + 0.00
reveals thatduringthis decade,the shareof articleswith one andtwo authors
decreased, while the share of articles with three authorsincreasedslightly
andthe shareof articleswithfourormoreauthorsincreasedquiteappreciably.
De Solla Price (1986) examinedsimilardatafrom 1900 to 1960, and Figure
2 compares his data with ours. In de Solla Price's data, the proportionof
articleswithtwo authorsgrewandtheproportionwiththreeauthorsincreased
rapidly.Between 1960 and 1980, the system seems to have madea transition.
Now, the proportionof articles with two authorsis declining, and soon the
proportionwith three authorswill probablybegin to decline. In the future,
only the proportionof articles with four or more authorswill continue to
grow, graduallyreplacingarticleswith threeor fewer authors.8
De Solla Price's (1986) data indicatethatthe need to combine skills and
labor in scientific work grew throughoutthis century.By 1950, the lone
investigator was the exception in chemistry,accounting for less than 50
percentof publishedarticles.Ourdataindicatethatmore recently,a second
transitionoccurred,a transitionin the breadthrequiredto producea piece of
knowledge. Today,two people and three people are less and less likely to
have all the skills, equipment,and materialrequiredin modern scientific
research.By the turnof the century,the currenttrendssuggest that four or
more authorswill be needed to producepublishedresearch.
The second graphin Figure 1 displays the frequencydistributionof the
number of domestic institutionsper article. We can see that the share of
articlesproducedby one institutionis declining,while the shareproducedby
two or more institutionsis on the rise. The final graphdisplaysthe frequency
distributionof foreign countriesper article.The shareof articleslisting one
foreigncountryincreasedby about 10 percentover the decade.The sharesof
articles with two or more countries increased by 2 percent. Both trends
suggest thatproducingresearchincreasinglydemandsnotjust thatindividuals pool resourcesbut that they do so across institutionaland even national
boundaries.
Hicks and Katz / Where is Science Going?
391
% shareof articles
40
T
- - - 1981
-?-
1991
Maximumin 1981- 97 =:
10
Maximumin 1991 - 521=
.,.
0
1
2
3
4
5
7
6
/ article
authors
9
8
10
11
12
: of articles
15
60
30
5
0
S
2 34
/ article
institutions
domestic
6
1
2
3
4
/ article
~~~~~~~~~~~~countries
Frequency distributions of authors, domestic institutions, and countries per article, 1981 and 1991.
NOTE: Lines plotting the data in the intervening years proceed smoothly from the 1981
to the 1991 position. The values plotted add to 100 in the case of authors and domestic
institutions because every article has at least one author and one domestic institution.
However, not every article has a foreign coauthor. The shares of countries per article
add to 14 percent in 1981 and 23 percent in 1991. The share of single-author,
single-institution, and single-country articles are decreasing. The share of multipleauthor, multiple-institution,and multiple-countryarticles is increasing. The exception is
the share of articles with two authors, which is decreasing.
Figure 1.
The resources of different institutions can sometimes be combined by a
single individual. This happens when a researcher moves and puts on the
article the name of the institution at which the work was begun and at which
it was finished or written up. It also happens when individuals hold joint
appointments (Katz and Martin in press). In most instances, these data do not
allow us to detect this activity. However, on 2.1 percent of U.K. articles, the
392
Science, Technology,& HumanValues
O
1900
*
I '"
1910
192
:-
1930
1940
1950
....
19O0
1f 70
1980
1990
Incidenceof MultipleAuthorshipas a Functionof
Date
Figure 2. Comparison of de Solla Price and Science Policy Research Unit data
on number of authors per article.
SOURCE: Figure for 1900 to 1960 from Little Science, Big Science and Beyond by
Derekde Solla Price.? 1986 by ColumbiaUniversityPress. Reprintedwithpermission
of the publisher.(p. 78)
number of institutions listed on an article exceeds the number of authors,
revealing either mobility or joint appointment. The percentage was steady
over the decade, although in 1991 it rose to 2.5 percent. Although we cannot
measure this activity precisely using these indicators, we should not forget
its existence. Temporary and enduring links between institutions are forged
both through collaborations between their employees and through joint
appointments and job mobility.
393
and Katz / Where is Science Going?
Hicks
Numberof articles
25000
noncollaborative
0 v v
~-
20000 -
oraiv
,.-"
collaborative
O.-.
15000 .-0--0'
10000-.0.
5000 -
0
1981
80 T
70
1983
1985
1989
1987
1991
1993
1995
1997
1999
2001
ofarticles
Percentage
o
-o0000000000000noncollaborative
60 50-
0.
40 30,
-o-- o- ..
'?
. o'----"
collaborative
0.-'0''0"
20t
0 I
1981
Figure 3.
I
1983
I
1985
}
1987
I
1989
}
1991
t
:
I
t
t
1993
1995
1997
1999
2001
Collaboration.
NOTE:The number and percentage share of collaborative and noncollaborative articles
inthe years 1981-91 are plotted with open circles. Collaborative articles are those listing
more than one institutional address. Collaborations between departments in one
university or between different sites of one company are counted as noncollaborative.
Linear regressions through the data are carried forward to the tum of the century when,
for the first time, the number of articles produced by researchers collaborating across
institutionalboundaries will exceed those produced by researchers working in a single
institution.Collaboration across geographical and institutionalbarriers will become the
rule, not the exception.
We next examine collaboration more directly by analyzing trends in the
numbers of collaborative and noncollaborative articles. Institutionally collaborative articles are defined as those listing addresses from different
institutions. Articles listing different department names within one institution
or different sites of one company are not considered to involve an institutional
394
Science, Technology,& HumanValues
collaboration.The growth in collaborativepublishing and the associated
slight decline in noncollaborativework are displayed in Figure 3. Both
numbersand shares of articles are displayed for each year, and each set of
data is accompaniedby a regressionline that is carriedforwardto identify
the point at which the lines will cross. Duringthe 1980s, the numberof U.K.
articlespublishedby authorslocated at one institution(the noncollaborative
articles)declined slightly,while the numberpublishedby authorsworkingat
more than one institutionrose steadily.These trendssuggest that the skills
and instrumentslocated at one institution are less and less adequate to
produceknowledge, and so researchersin differentinstitutionsincreasingly
combineresourcesto produceandpublishresearch.If these trendscontinue,
the shareof collaborativearticleswill exceed thatof noncollaborativearticles
sometime aroundthe turn of the century.Thus, as Britainenters the next
millennium,its researchsystem will undergoa transitionin which research
collaborationamong geographicallyseparatedinstitutionswill become the
normalway of conductingresearch-the rule,not the exception.Both Ziman
and Gibbons et al. discuss the policy and managementimplicationsof this
transition.
Internationalcollaborationis often singled out for special mention.It has
been a concern of recent EU science policies and of bibliometricanalysis
(Schubertand Braun 1990). Ziman arguesthatthe increasedcost of certain
instruments,the increased scope of many problems, the global reach of
research-intensivemultinationalcompanies, and increasedtravel and communication are combining to make the scientific community even more
transnational-researchhavingalwaysbeen a moreinternationalpursuitthan
most. Gibbons et al. also believe that increasinginternationalcollaboration
arises from forces intrinsicto the researchprocess, forces operatingover a
very long time scale. Both connect rising internationalcollaboration to
increasedtraveland communication.
In an attemptto illustratethis thesis, we examineda longer time series of
collaborative data (Katz and Hicks 1995b). The first graph in Figure 4
displays the percentage of U.K. and U.S. articles with an international
collaboratorfrom 1975 to 1991. In this figure, our data are combined with
those published in the U.S. National Science Board's (1993) Science and
EngineeringIndicatorsto providea longertime series.Internationalcollaborationhas been increasingfor fifteen years, both in the United States and in
the United Kingdom. Figure 4 also compares the trends in international
collaborationwith trends in travel and communication-specifically with
numberof internationalscheduled air passengers and numberof outgoing
internationaltelephone calls. All three graphs exhibit long-term growth
originatingin the mid-1960s. Growth in collaborationis the most linear,
395
Hicks and Katz / Whereis Science Going?
25
International
scientificcollaboration
UK
20
. I-
15 -
-
US
6^^
_10 -
0~
o
-~~~
. .
.....
.............
:~ OO?
:~
............
Articles,notesandreviews
O16
1963
3
1966
1969
1972
1975
1978
1981
1984
1987
1990
1990
2
40
.1 40
E
C
30
2
20
X 10
Z
,
U
1'963
1966
1969
1972
1975
1978
1981
1984
1987
1966
1969
1972
1975
1978
1981
1984
1987
1000
s800
o
4@400
z
200
1963
1990
Figure 4. Growth in international scientific collaboration, international scheduled air traffic, and outgoing international telephone traffic, United
States and United Kingdom.
SOURCES: U.S. National Science Board (1993), International Civil Aviation Authority
(1975-1993), InternationalTelecommunication Union (1975, 1977-1979, 1990)
growth in air travel is the most subject to economic cycles, and growth in
telephone calls is the closest to exponential. Between 1981-83 and 1989-91,
396
& HumanValues
Science,Technology,
the number of articles produced in internationalscientific collaboration
increased by 74 percent; the number of internationalairline passengers
carriedfrom the United Kingdomincreasedby 87 percent,and the number
of internationaltelephonecalls from the United Kingdomincreasedby 138
percent.The graphssuggest thatZimanand Gibbonset al. are rightto place
internationalcollaborationin a broadercontext.
The increasesin collaborationsuggest thatthe competencies, skills, and
materialresourcescombinedto producean advancein knowledgearegrowing graduallybut inexorablyover time. Some types of researchare already
well known for being organized in large internationalconsortia-highenergy physics, space, oceanography,polar, and other environmentalsciences. If thesetrendscontinue,however,producingscientificandtechnological knowledge of any type will come to be seen as a matterof coordinating
dispersedgroupsof people-though groupsin most areasmay neverbecome
as large as those in the aforementioned"consortiasciences."
Wemightthenwonderwhatwill be theendpointin this evolution.Perhaps
the system will asymptoticallyapproachcomplete connection (i.e., every
institution collaborates with every other institution). Or perhaps it will
approach some point in between completely connected and completely
disconnected. If we accept some of the new notions that the amount of
complexity in a system is at a minimumboth when the institutionshave no
connections (no collaborations)and when each is connectedto every other
institution(completelycollaborative),then perhapssomewherebetween the
extremes the system is most complex with institutionsconnected yet still
flexible enoughto respondto externalchanges(Gell-Mann1994;Crutchfield
1994). Whereis thatpoint?And is the Britishscience system approachingit?
A longer time series of datawould be needed to addressthese questions.
Concentration or Dispersion of Research Production?
Another way of interpretingthe increase in collaboration is that the
production of knowledge is becoming more socially dispersed. In other
words,a "piece"of knowledgewill in the futurebe producedby morepeople
in more locations.As the productionof pieces of knowledge becomes more
dispersed,will the capabilityto produceknowledge become more dispersed
too? The authorsdisagree on this point. Gibbons et al. point to increasing
dispersionin the system. Accordingto theiranalysis, thereis
an increase in the numberof potentialsites where knowledge can be created;
no longer only universitiesandcolleges, but nonuniversityinstitutes,research
Hicks and Katz / Whereis Science Going?
397
consulthink-tanks,
centres,governmentagencies,industriallaboratories,
tancies,in theirinteraction.
(Gibbonset al. 1994,6)
In contrast,Ziman sees the researchsystem as becoming more concentrated.He points to very powerfulforces of competitionbased on excellence
that are endogenous to science and that lead to concentrationof resources
over time. In addition,he notes thatcloser managementof researchleads to
even more concentrationto achieve administrativeefficiency, economies of
scale, and division of labor."Selectivity"is, in his view, increasinglypartof
managing steady state science: "What selectivity has come to mean in
practice is a systematic policy of concentratingresearch activity into a
smallernumberof more specializedunits"(Ziman 1994, 156).
We can examine bibliometricallywhether the U.K. science system is
becoming more dispersed,as Gibbonset al. propose, or more concentrated,
as Ziman believes. Our data permit us to look at several facets of the
concentration/dispersionphenomenon.We can analyze the types of institutions publishingresearchto see whethera more variedset of institutionsare
participatingin the system. We can also examine the numberof institutions
publishing to look for increases. Finally, we can look at concentration
measuresusedby economiststo see whetherresearchproductionis becoming
more evenly distributedamong institutions.
We begin by examiningthe types of U.K. institutionsthatpublish,orrather
the types of institutionsthat publish articles, notes, or reviews in journals
included in the SCI. These 3,000 or so journals were selected in the first
instancebecause they have a high internationalimpact.Indeed,coverage of
the database has been criticized because the criteria for the inclusion of
second-rankjournalsareinconsistentandappliedfields arenot well covered
(EuropeanCommission 1994, 33-34). In addition,we countedonly articles,
notes, andreviews becausethey aremost likely to reportsubstantialresearch
results and be peer reviewed. Discussions, letters, editorials, and meeting
abstractshave been excluded.
Therefore,theSCI as analyzedhererepresentsinternationalpeer-reviewed
science-the domain of academics.Why then do 60 percentof articles list
the addresses of noneducationalinstitutions?Why do hospitals and firms
account for 63 percentof the institutionsthat averagedone or more articles
per year? Why did ICI, Wellcome, SmithKline Beecham, and BT each
contributemore than 1,000 articlesduringthe 1980s? Because of collaboration, about 60 percent of articles also list the address of an educational
institution.9Nevertheless, the fact that so much international-level,peerreviewed, scientific knowledge in the United Kingdomis producedoutside
the universitysector,or in collaborationwith institutionsotherthanuniver-
398
Science, Technology,& HumanValues
Table 6. Selected U.K. Institutions That Published
Articles from 1981 to 1991
Animal Health Trust
Atkinson Morleys Hospital
British Ceramic Research Ltd.
British Railway Board
BritishTrust Ornithology
Forensic Science Service
Forestry Commission
Greenwich District Hospital
50 to 150
Malaysian Rubber Producers' Research Association
Meat and Livestock Commission
National Museum of Wales
Nature Conservancy Council
Organon Research Labs Ltd.
Royal Botanical Gardens, Edinburgh
Royal Museum Scotland
Royal Society for the Protection of Birds
sities, suggests that universities do not have a monopoly on "academic"
research.Medical institutions,industriallaboratories,researchcouncil and
othergovernmentlaboratories,and nonprofitinstitutescollectively seem to
be as importantas universitiesin the modernU.K. researchsystem. Many of
these institutions,such as local councils and police forces, are not normally
thought of as contributingto Britain's scientific researchoutput. Table 6
illustratesthis institutionalvariety by listing a selection of institutionsnot
known for their researchoutput(not universities,polytechnics, or research
council laboratories)thatpublishedbetween 50 and 150 articlesduringthe
decade.
Analyses of that part of the researchsystem publishing in international
refereedscientificjournalsmust take into accountthe varietyin the system.
An exclusive focus on researchin highereducationis a distortion.
Gibbonset al. not only pointto this varietyin the researchsystem but also
claim thatit has been increasing.To investigatethis, we can ask whetherthe
numberof participantsin the researchsystem has been increasing.Figure 5
displays for each sector three-year moving averages of the number of
institutionspublishingfor each year 1983 to 1991. In the graph,publishing
institutions are classified into four categories according to whether they
publishedon average 1 articleper year,2 to 10 articles, 11 to 100 articles,or
more than 100. (Only Oxford and Cambridgepublished more than 1,000
articles per year.) The graph reveals that in this area, sectors have very
different characteristics.For example, many hospitals or companies published just 1 articleor less per year, while most universitiespublishedmore
than 100 articlesper year.m?
The graph shows that the numberof institutionsat which publications
originateincreasedduringthe 1980s in all sectors but one. The numbersat
the beginningandend of each sector'sgraphreportthe numberof institutions
in the first and last years. If we takethe differencebetweenthese numbersas
HicksandKatz/ Whereis ScienceGoing? 399
Averagearticlesperyear
of
Number
institutions
>100
11to?100
2tolO0
hospitals
.
600 5S~
*
_
_
=
^^
! tc=:8
iI
I 69
L
1BH|g
68
69
|^
^
5
ia
I r-^-
..1
Figure 5. Numberof publishing institutions.
NOTE:Three-yearmovingaverages ofthe numberofinstitutionspublishing.Institutions
are classified by average number of articles per year. Only Oxfordand Cambridge
publishedmorethan 1,000 articlesperyear.The numbersindicatethe totalnumberof
institutionsin the sector publishingin the firstand last three-yearperiods.The number
of institutionsincreased in all sectors except research councils.The universitysector
consists mostly of large publishers;research councils, of medium-sizedpublishers;
hospitals,of smallpublishers;and industry,of companiespublishingonlyone articleper
yearon average.
a measure of the size of the change, the numberof firms increased by 32
percent,~land the numberof hospitalsincreasedby 11 percent.The number
of nonprofitinstitutions,level throughmost of the decade, increasedin the
lasttwo years by 19 percent.12The numberof polytechnicsincreasedalso by
12percent,though this correspondsto just four institutions.The numberof
governmentinstitutionsvaried over the decade, though the numberin the
final year was 11 percent higher than in the first year.13The number of
universitiesremainedessentially constant.The exception to the increase is
theresearchcouncil sector in which numbersdecreasedby 13 percentdue in
largepart to consolidationby the Agricultureand Food ResearchCouncil.
Theincreasein the numberof researchcouncil laboratoriesproducingmore
than100 articlesper year (the top band)is consistentwith concentration.
Theincreasingnumberof institutionshousing authorsof journalarticles
lends supportto the idea that researchproductionis becoming more dispersed.Only the decline in numberof researchcouncil laboratorieslends
supportto the idea thatmore managementof researchis leading to consoli-
400
Science, Technology,& HumanValues
dation and concentration.The weight of evidence favoringdispersionreinforces the point thatacademicresearchaccountsfor only half of the research
system in the United Kingdom today. In fact, it forms the static half. The
numberof institutionsof other types that producejournal articles, such as
companies and hospitals,has grown.
The number of institutions is not by itself an adequate measure of
concentration,however.Concentrationhas two dimensions:both numberof
organizationsand the inequalityof size among them play a role. The distribution of researchproductionacross institutionshas always been uneven,
with a few institutionsproducinga greatdeal and a large numberproducing
little. WhereasZimanpoints to the forces (traditionalandinternal-new and
external)that producethis distribution,the analysis by Gibbons et al. suggests thatthe distributionmight be becoming more even.
Figure5 displays the numberof institutionshousing authorsof scientific
articles, qualitativelyconveying differences in how much they publish. To
analyzeconcentrationmorequantitativelyandto examinewhethertherewere
any changesover the decade,we will examinemeasuresof concentration(see
Table 7). As there are various measuresof concentration,none ideal in all
circumstances,we presenttwo such measureshere, bothof which reflect the
two dimensions of concentration(Davies et al. 1988, 79-86). The first
measureis the numberof firmspublishing25 percent,50 percent,75 percent,
and 100 percentof the articles,which is easy to understandand conveys the
natureof the tail in the distributionfor each sector.The second measure is
the Herfindahlindex, which, being a single number,makes it possible to
assess change in concentrationover time. This index is calculatedby summing the squaresof each institution'sshare of the sector's publication(its
maximumvalue is 1). For example, if therewere two institutionsin a sector
producing95 percentand 5 percentof the articles,respectively,the Herfindahl index would be 0.952+ 0.052= 0.905. In a sectorof size n, concentration
reachesa minimumwhen all institutionspublishthe same numberof articles.
At this minimum, the Herfindahl index would equal l/n, where n is the
numberof institutions.By reversingthis logic, each sectorcan be seen to be
as concentratedas a sector of 1/Herfindahlindex number of equal-sized
institutions.This numberof institutionsis reportedin the last line of the
tables. There are three panels in the table. The first two reporta three-year
average, in 1981-83 and 1989-91, respectively.The final panel reportsthe
difference between the figures in the first two panels. The columns are
orderedby how concentratedthe sectors were in the 1989-91 period.
The tables indicate that the sectors vary somewhat in their degree of
concentration.In the first part of the decade, government was the most
concentrated;in the latterpart, the nonprofitsector was the most concen-
Table 7.
Concentration Measures
Nonprofit
1981-83
Percentage of articles
25
50
75
100
Herfindahlindex
Sector size equivalent
1989-91
Percentage of articles
25
50
75
100
Herfindahlindex
Sector size equivalent
>
?
Difference between 1989-91
and 1981-83
Percentage of articles
25
50
75
100
Herfindahlindex
Sector size equivalent
SHA&
Resea
BPG Government Industry Polyte
1
3
10
109
0.115
2
4
7
17
0.108
9
9
2
3
7
86
0.121
8
1
2
9
149
0.151
7
2
4
8
18
0.098
10
2
4
8
91
0.095
11
0
-1
0
0
0
-1
40
0.036
-2
1
1
-0.010
1
1
1
5
-0.026
2
3
12
52
723
0.033
30
1
3
0.04
2
3
11
1
62
992
0.036
28
2
3
0.03
2
0
-1
10
269
0.003
-3
-0.013
402
Science, Technology,& HumanValues
trated. Hospitals are the least concentratedsector. Sector concentrations
changedsomewhatover the decade.Five sectorsbecamemoreconcentrated:
researchcouncils, industry,nonprofit,university,and hospital;threebecame
less concentrated:polytechnics,government,and SHA andBPG. Thus there
is no uniformtrendtowardconcentrationor dispersion.Some changes are
clearly relatedto governmentpolicy, those in the polytechnicsand research
councils in particular.However, even government policy did not have a
uniformeffect, producingdispersionamongthe polytechnicsandconcentration among researchcouncil laboratories.
Conclusion
As the economy becomes more knowledge based, workers, firms, and
governmentsare adapting.It is hardlysurprisingthen that the knowledgeproducingscientific system is changingas well. The proportionof research
that is interdisciplinaryis increasing,collaboration-domestic and international-is rising steadily,and more institutionsare producingresearcharticles. Researchprojectscombinean everbroaderrangeof skills andresources,
indicated by increasinginterdisciplinarityand collaborationbetween individuals,institutions,andcountries.By the turnof the century,one-half of all
articles will probablybe producedby four or more authorscollaborating
across institutionalboundaries.
Science policy mustalso changeand,in fact, has alreadybegunto change.
Both Ziman and Gibbonset al. point to changes underway and those likely
in the future.Knowinghow to fund,manage,facilitate,andconductcollaborative researchwill become core scientific and policy competencies in the
next century.Evaluationmethodsmust adapt.Methodsbased on examining
a unit's "own" research output in comparison with others will not work
when individuals, groups, departments,or institutions do not have their
"own"research output because more than 50 percent of their research is
collaborative.
Knowledge-producinginstitutionsare more heterogeneousthan is perhapscommonlyrealized,andthis varietyis increasingas more nonacademic
institutionsbegin to publishscientificarticles.To the extentthatarticlesfrom
small,new publishersreflectin-houseresearch(andwe hope to establishthis
in futureresearch),such organizationsmay mimic smallbiotechnologyfirms
in accommodatingresearchoutside academia.Even today, academics have
no monopolyon knowledgeproduction.Moreflexible, focused, andcontextsensitive institutionsmay mountan increasinglyvisible challenge to academia as theirnumbersincrease.
Hicks and Katz / Whereis Science Going?
403
Literature-basedanalysis paintsa pictureof gradualchange thatis somewhat at odds with impressions gleaned from qualitative work. Ziman in
particularsuggests dramaticchange;Gibbonset al. do to a lesser extent. Of
course, Ziman discusses managerialchanges, which may well be more
dramaticthanthe changes we examine.
Bibliometric data can reveal only so much, however, and we should
not forget those aspects that are invisible to bibliometricanalysis but are
addressedby Ziman and Gibbons et al. Like Gibbons et al., we examine
science at the systemic level. Gibbonset al. recognize the hardshipsuffered
by individualscaught up in the ongoing changes but do not focus on this.
Zimanuses the systemic changes more as the setting for a discussion of the
hardshipfaced by scientists and the difficulties of an academiccareerin the
1990s. Bibliometric analysis provides no indication of the stress felt by
individualsin the system. However,we would say thatone cannotgeneralize
fromindividualhardshipto concludethatthe systemis falling apart.Difficult
as the adjustmentsare for the scientists affected, there is no evidence here
thatthe British science system is disintegratingor in decline.
The authors also make a general argumentabout scientific research,
whereaswe examine only U.K. research.Data for othercountrieswill look
different; for example, a higher percentageof articles may list academic
addresses.However, if Ziman and Gibbonset al. are correct,the trendswill
be the same. We are working with groups in other countries to produce
comparativedatato generalizeour empiricalargument.
Any analysis of dynamic change in a system must be grounded in an
understandingof the "energy"inputsto the system. In science, therearetwo
such inputs:people andmoney.Zimananalyzeshow the systemis responding
to reducedgrowthin the money allocatedit. He has a very clear graspof the
importanceof funding, while Gibbons et al. give much less weight to this
factor.This may also explain the differencebetween theirpredictionsabout
researchconcentration.The increasingdegree of concentrationamong researchcouncil institutionsindicatesthatfundingmust be given prominence
in any systemic analysis of science. On the other hand, Gibbons et al.
emphasizethe huge post-WorldWarII growthin the numberof college-educated and Ph.D.-trainedpeople in Organizationfor Economic Cooperation
and Development (OECD) countries.They realizedthat this growthhas led
to researchcompetencebeing distributedmuchmorewidely throughsociety,
no longer confined to universities.Ouranalysis confirmsthe significance of
this trend.
Ourdatado not deal with instrumentation.Gibbonset al. discuss it briefly
and in a somewhat abstractway. Ziman gives it a prominentplace in his
404
Science, Technology,& HumanValues
analysis. We believe that the scientific community and scientific instrumentation are coevolving systems. If comprehensive data about the performance,
cost, and scale of instruments and the connections between them were
available, they might explain some of the trends we have found.
Finally, we should not forget the invisible years; although a decade of data
seemed like a lot at the beginning of this project, it now seems like just the
beginning. To some extent, the conflicting ideas of Ziman and Gibbons et al.
can be put in perspective by considering the time periods they cover. Ziman
focuses on changes over the past decade or so, the period for which we have
bibliometric data. Gibbons et al. analyze the longer term, approximately the
last fifty years. Over this longer period, we would certainly see entry into
publishing of large numbers of institutions in each sector. The postwar period
was a time of institution building, before the steady state was reached. New
universities were established during this period, for example. We might
hypothesize that as the institution-building, expansive phase ended in the late
1960s, networking and collaboration began to increase. Only extended time
series bibliometric data going back thirty years or so would allow us to
investigate and to understand fully the complexity and order in the U.K.
science system.
Notes
1. The reasonswhy companiespublisharticlesare analyzedin Hicks (1995).
2. "Hospitalpublishing"meansthose articleslistinga hospitaladdress.Since manyarticles
list addressesof institutionsin more than one sector, the percentageshare figures add to more
than 100 percent.Because the rateof collaborationis increasing(see below), the extent of what
we mightterm"overlappingshares"is increasing,and it is conceivablethatevery sector's share
of publicationscould have increased.
3. These interfieldcategoriescontainthe articleswe call interdisciplinary
in the next section.
4. Indeed,in the early 1980s, ISIjournalswere classified into fourcategoriesbased on how
basic or appliedwas theresearchtheyreported.Unfortunately,thisscheme has not been updated.
Therefore,we could not use it here.
5. This trend,and the growthin interdisciplinaritydiscussed in the next section, could be
caused by expandedcoverage of the SCI if many new journals were added and they were all
applied.However,duringthe 1980s, the coverageof the databasewas relativelystable. In 1981,
it indexed 3,068 journals(or "sourcepublications");in 1991, this hadbecome 3,213 journals(a
5 percent difference). The maximumnumberof journalswere indexed in 1983-3,327. Thus
expandedcoverage of the SCI does not confoundthe trendswe report.Nevertheless, the SCI
grew. Even with a relatively stable numberof sourcejournals,the numberof articlesgrew by
12 percent(from 387,000 to 434,000).
6. ScottCunningham(Science Policy ResearchUnit, Universityof Sussex, D.Phil.research
in progress)is developing techniquesfor classifying large numbersof articleson an individual
basis.
Hicks and Katz / Whereis Science Going?
405
7. In the interveningyears, the curves move smoothly from the 1981 to the 1991 position.
8. Analysis by the ISI reveals that since 1989, the numberof articles with more than 50
authorshas increasedsharply.Similartrendsare visible in numbersof articleswith more than
100, 200, and 500 authors(ISI 1995).
9. If we divide each sector's shareof publicationsby the sum of sectorshares,the resulting
"normalized"shareswill add to 100. In this case, educationalinstitutionsaccountfor about53
percentof U.K. output,and noneducationalinstitutionsaccountfor 47 percent.
10. Those that do not are, in general, oddities, such as the London Business School,
publishinga few articlesin journalsindexed in the SCI.
11. Companynames were unified to the 1989 Who Owns Whom, and thus the numberof
parentcompaniesare counted.
12. The following nonprofitinstitutionspublishedno articlesbetween 1981 and 1983 but
published five or more articles between 1989 and 1991: British Heart Foundation,British
InstituteReflect Profiling Syndicate, British Society HorticultureResearch,Essential Rights,
Fund for the Replacementof Animal Medical Experimentation,National Society for Epilepsy
ResearchGroup,PainRelief Foundation,QuadrantResearchFoundation,ThrombosisResearch
Institute,Wynn Instituteof MetabolismResearch.
13. Laboratoriesprivatizedduringthe decadewere categorizedas industryduringthe whole
period. If they had been moved to industryas they were privatized,the numberof government
instituteswould have declinednoticeablyand the numberof companiesrisen even more.
References
Crutchfield,J. P. 1994.The calculiof emergence:Computation,dynamicsandinduction.Physica
D 75:11-54.
Davies, Stephen,Bruce Lyons, Huw Dixon, and Paul Geroski. 1988. Economics of industrial
organisation.London:Longman.
De Solla Price, Derek. 1986. Little science, big science and beyond. New York: Columbia
UniversityPress.
EuropeanCommission. 1994. The Europeanreporton science and technologyindicators1994.
EUR 15897. Luxembourg:Office for Official Publicationsof the EuropeanCommunities.
Gell-Mann,Murray.1994. The quarkand the jaguar: Adventuresin the simple and complex.
Boston: Little, Brown.
Gibbons, Michael, Camille Limoges, Helga Nowotny, Simon Schwartzman,Peter Scott, and
MartinTrow. 1994. The new productionof knowledge.London:Sage.
Hicks, Diana. 1995. Published papers, tacit competencies and corporatemanagementof the
public/privatecharacterof knowledge. Industrialand CorporateChange4:401-24.
Institutefor Scientific Information(ISI). 1995. Really big science: Multiauthorpapersmultiplying in 1990s. Science Watch6 (April): 1-2.
InternationalCivil Aviation AuthorityICAO StatisticalYearbook. 1973-1993. Civil Aviation
Statisticsof the World.Montreal:InternationalCivil AviationOrganisation.
InternationalTelecommunicationUnion. 1975, 1977-1979, 1990. Yearbookof CommonTelecommunicationStatistics.Geneva: InternationalTelecommunicationUnion.
Katz, J. Sylvan, and Diana Hicks. 1995a. A classificationof interdisciplinaryjournals:A new
approach.In Proceedingsof 5th internationalconferenceon scientometricsand infometrics,
edited by Michael E. D. Kounig and AbrahamBookstein, 245-54. Medford,NJ: Learned
InformationInc.
406
Science, Technology,& HumanValues
1. 995b. Questionsof collaboration.Nature375 (11 May): 99.
Katz, J. Sylvan, DianaHicks, MargaretSharp,andBen R. Martin.1995. The changingshape of
Britishscience. STEEPSpecial ReportNo. 3. Brighton,UK: Science Policy ResearchUnit.
Katz, J. Sylvan, and Ben Martin.In press. Whatis researchcollaboration?ResearchPolicy.
Leclerc, M., Y. Okubo, L. Frigoletto,and J.-F.Miquel. 1992. Scientific co-operationbetween
Canadaand the EuropeanCommunity.Science and Public Policy 19:15-24.
Luukkonen, Terttu, Olle Persson, and Gunnar Sivertsen. 1992. Understandingpatterns of
internationalscientific collaboration.Science, Technology,& HumanValues17:101-26.
Narin,F., and E. S. Whitlow. 1990. Measurementof scientific co-operationand coauthorshipin
CEC-relatedareasof science. Vol. 1 of Reportto the Commissionof the EuropeanCommunities. CherryHill, NJ; CHI Research.
Schubert,A., and T. Braun. 1990. Internationalcollaborationin the sciences. Scientometrics
19:3-10.
U.S. National Science Board. 1993. Science & Engineering Indicators-1993. NSB 93-1.
Washington,DC: U.S. GovernmentPrintingOffice.
Ziman, John. 1994. Prometheusbound: Science in a dynamicsteady state. Cambridge,UK:
CambridgeUniversityPress.
Diana M. Hicksis a senior lecturerat the Science Policy ResearchUnit.She has recently
published examinationsof the relationshipbetween tacit and published knowledge(in
Industrialand CorporateChange),university-industry
researchlinksin Japan(in Policy
Sciences), and Japanese-Europeancorporate researchnetworks(in ResearchPolicy).
Her currentresearch interestsinclude understandingthe evolutionof science systems,
particularlythe contributionof companiesto them.
J. SylvanKatzis a researchfellow at the Science Policy ResearchUnit.He has recently
publishedexaminationsof the relationshipbetweenthe evolutionof science systemsand
governmentpolicy (in Nature), the relationshipbetweengeographical separation and
university-universitycollaborationwithin three countries (in Scientometrics),and the
informationcarriedbyfractal distributions(in SportsMedicine,Training,andRehabilitation). His current research interests include techniquesfor analysis of large text
databases, theeffectof policy on researchsystems,and complexityand self-organization
in science systems.