本研究提出的方法特別關注學科之間的相互關係，提供了一個所有科學學科的總體概述。根據基於ISI(Institute for Science Information)的一系列多重學科分類指標，測量學科和研究領域的跨學科性，進而建立一個初步的學科和研究領域型態學(typology)。以研究領域和類別上的連結數量描述它們的相關類別數量及接近或遙遠的類別，多樣性和連結強度等方面的測量。
跨學科性問題的研究可以利用多種不同方法，例如通過訪談和調查(Hargens, 1986; Palmer,1999)，或是對於高等教育系統的組織分析和研究團隊的實證分析等(Sanz et al., 2001)。也可以利用書目計量學方法，特別是利用詞語、作者或參考文獻做為共現分析資訊，確定不同子領域之間的結構關係，並將其呈現為圖形表示的“科學地圖”(例如Tijssen, 1992; Small, 1999; Weingart & Stehr, 1999)。還有利用來自不同學科作者或中心之間的合作((Qiu, 1992; Qin et al., 1997; Bordons et al., 1999)，以及著重參考文獻或引用在類別上的分布(Porter & Chubin, 1985; Urata, 1990; Tomov & Mutafov, 1996; Bourke & Butler, 1998; Van Leeuwen & Tijssen, 2000; Van Raan & Van Leeuwen, 2002; Rinia et al., 2002)。上述的研究中通常將學科視為是期刊的集合(journal sets)，“跨界”(boundary crossing)的研究者便被認為是那些在本身學科以外的期刊上發表的作者。在Pierce(1999)關於社會學和政治學的研究中，發現跨界作者偏向來自鄰近學科，並從論文的被引用率可以證實，這些作者成功地實現跨學科信息傳遞。Steele 和 Stier (2000)對森林學的研究則發現跨學科的文章(意即其引用的參考文獻較多元)比集中在學科內的文章有較高的引用率，因此他們認為科學家會使用其他學科的文獻來增加其研究的影響力。宏觀層面的跨學科性分析主要處理學術領域之間的結構關係，例如引用流向(citation flow)提供科學地圖的圖形表示研究（例如參見Small，1999）或通過不同領域間學者的遷移(Hargens, 1986)。關於學科之間知識交換，先前的研究已經使用文獻中的跨學科引用進行分析(Van Leeuwen & Tijssen, 2000; Rinia et al., 2002; NSF, 2002)，這些研究都利用期刊分類的類別做為文章歸屬的子領域(subfields)。Rinia et al. (2002)提供了對領域之間關係的有趣洞察，並根據不同的跨學科性度量提出了領域排序系統(field ranking system)。
本研究中根據期刊分配到一個以上主題類別的特徵，也就是多重指派(Multi-assignation)，來檢驗科學的跨學科性，其根據的假設是指派為一個以上主題的期刊應該比單一指派的期刊更具跨學科性。本研究認為將期刊分配到多個類別表明學科之間存在認知聯繫，這可能進一步導致跨學科研究，並且可以追溯到一段時間的演變。以下是本研究的問題：首先，根據其跨學科性質描述類別，並建立不同類別的型態。其次，將研究領域(research areas)描述為它們類別的集合，探索這種自下而上的方法的可能性，並且以獲得領域的類型學為目的。最後，假定最近創建的類別由原先學科的專殊化(specialization)或混合而來，因此它們比較舊的類別更具跨學科性，根據期刊多重指派的指標，新增和增長最快的類別是否具有較高跨學科性。
本研究利用1996年ISI的SCI、SSCI和A&HCI資料庫上的8000種期刊，共有224個類別，每一個期刊至少被指派一個類別，最多則為五個類別。每一個類別包含的期刊，從4種到284種，分布相當偏斜(中位數為38，平均約為50)。期刊類別的指派是根據期刊的內容以及對其引用和被引用的模式進行分析所歸納出來的結果。本研究所使用的指標如下：
1. 每個類別的多重指派期刊的百分比，百分比愈高表示該類別的跨學科性愈高。
2. 多重指派為同一領域內的類別或為不同領域間的類別百分比。
3. 關係的多樣性(diversity of relationships)，由特定類別建立的不同連結的數量，即與此類別有共享期刊的不同類別的數量計算得到。
4. 兩個給定類別之間建立的關係的強度，類別A和類別B之間的關係的強度被計算為A和B共享的期刊數量除以A中的期刊數量的平方根乘以B中的期刊數量之間的比率，也就是類別A和類別B的Salton指標(Salton index, Salton & McGill, 1983)。
研究中，將區域和類別分群以獲得不同的型態，對出現的分群結果中眾所周知的跨學科類別的位置進行分析，做為最終型態的驗證標準。並且利用Noma (1986)對期刊研究層級的研究結果判斷各類別的基礎或應用科學。最後，分析了過去幾年ISI類別的演變情況，也利用多重指派指標對過去15年中最新加入ISI分類體系中的學科類別進行識別和描述，以檢驗新興學科具有更高跨學科性的假設。
平均上來說，每一個類別上有53%的期刊其主題為多重指派，看起來這個比例似乎相當高，但從知識結構的角度來看，似乎是合理的，因為學科的邊界是人為界定，不同學科的知識之間確實存在高度關係。在各類別上，期刊的多重指派比例有很大差異：人文領域的類別幾乎沒有多重指派的期刊，在某些較小的類別上，全部的期刊都是多重指派，較大並具有較高多重指派比例的類別包括生物心理學(Biological Psychology, 100%)、熱力學(Thermodynamics, 100%)、行為科學(Behavioral Sciences, 97%)、環境工程(Environmental Engineering, 96%)、儀器學(Instruments/Instrumentation, 96%)和醫學資訊學(Medical Informatics, 94%)等。
各領域的多重指派比例，以及多重指派為同一領域內的類別(internal)或為不同領域間的類別(external)的百分比如下
各領域期刊的多重指派比例，最高為Engineering/Technolgy(56.5%)和Biomedicine(56.4%)，最低則為Humanities(11.0%)，而且Humanities與其他領域差距很大。從期刊的多重指派比例來看，基礎科學(例如Biomedicine)和應用科學(例如Engineering/Technolgy)並沒有很大差異。
除了Social Sciences之外，其餘的領域多重指派為不同領域間的類別的比例較高，最多者為Chemistry、Physics和Mathematics，多重指派為同一領域內類別的比例較高者為Social Sciences、Humanities和Engineering/Technology。利用多重指派的比例和內/外部連結進行階層式叢集分析，其結果分為四群，Social Sciences與Humanities和其他領域有明顯不同，彼此也不相同，因此兩者各成一群。
平均而言，每個類別連結14個其他類別(平均為12，在1~49)之間，而且擁有期刊數愈多的類別，其連結愈多，例如：Biochemistry & Molecular Biology有284種期刊，共有49個連結；Environmental Sciences有120種期刊，共有46個連結； Psychology則有139 journals種期刊，有39個連結。在以期刊數調整後，各領域的連結分布，如下表所示：
以兩個類別之間的關係強度而言，具有較強關係的類別有Remote Sensing、Mathematical Psychology 、Physics/Particles & Fields以及Petroleum Engineering，其中Remote Sensing不但具有較強的關係強度，同時也具有較高的多樣性。下面是各領域具有的類別之間的關係強度：
根據各領域間的具有相同期刊的連結強度，利用多維尺度(multidimensional scaling)，呈現它們的相似性，
首先利用區辨分析(discriminant analysis)判別多重指派比例、外部多重指派比例、多樣性和連結強度等各種指標的重要性，結果只有具有低區辨力的多樣性被移除，然後根據其餘三種指標，對領域進行階層式集群分析(hierarchical cluster analysis)。結果如下：9個領域明現地分為兩群，Social Sciences和Humanities在一群，其餘的領域在另外一群。若是以更嚴格的標準來看，可以分為6群：第一群在各個指標上的結果都很高，可以說是最具有高跨領域性的領域，這群的成員包含Biomedicine和Engineering。第二群則有Physics和Chemistry，兩個領域都具有中高的多重指派比例和高外部多重指派比例，但只有中等的連結強度和多樣性。第三群包括Agriculture/Biology/Environmental Sciences和Clinical Medicine，四個指標都為中等程度。其餘三個領域各自成一群。Social Sciences有中等程度的跨領域性，但外部多重指派比例遠較內部多重指派比例來得低。Mathematics也是中等程度的跨領域性，但具有高外部多重指派比例以及低連結強度和多樣性。Humanities則是擁有最低的跨領域性。
進行類別的分群時，同樣利用區辨分析選擇多重指派的相關指標。在這個程序中，連結強度因為區辨力較低而被移除，因此根據其餘三種指標對219個類別(排除少於9種期刊的類別)進行K-平均(K-means)集群分析。在產生的結果中，第一群(cluster a)的57個類別具有最高的跨學科性，包括Thermodynamics、Biological Psychology以及Mathematical Psychology，許多類別有極高的連結多樣性，例如Biological Psychology連結其他10個類別，Medical Informatics則連結12個類別，在這群中許多類別直觀上便是跨學科，例如：Biochemical Research Methods、Biotechnology/Applied Microbiology、Environmental Sciences、Applied Physics、Applied Chemistry以及Medical Informatics等。根據多重指派比例，第二群(cluster b)的66個類別也具有較高的跨學科性，但它們的內部連結高於外部連結，而且多樣性和連結強度略低於第一群，內部連結較高顯示這些類別連結的其他類別大多屬於相同領域，例如Instrumentation連結的類別，與其同樣都屬於Engineering，Limnology連結其他的Agriculture領域下的類別，Transplantation則和Clinical Medicine領域的類別連結。第三群(cluster c)表現出中低水平的跨學科性，它們的多重指派比例較低，但是外部連結高於內部連結。這群共有55個類別，包括Psychology、Chemistry、Neurosciences以及Information Science/Library Science，Information Science雖然屬於Social Sciences領域，但和屬於Engineering領域Computer Science/Information System有關，所以它的外部連結比同屬於這一群的其他類別高。第四群(cluster d)的跨領域性最低，Humanities領域中約有70%的類別在這一群。
1981到1996年各領域中，以Engineering新增的期刊最多，平均每個類別增加154%，其次是Mathematics的134%，Physics的95%和Clinical Medicine的91%，低於平均值的有Biomedicine (50%)、Chemistry (57%)和Agriculture (75%)。
15年中新增的類別，共有38個，大部分是從舊的類別分離出來(例如Materials Science分離出Ceramics、Coverings、Biological Materials等)或是源自其他類別的混合(例如：Biotechnology、
Infectious Diseases、Transplantation)。擁有最多新增類別的領域是Engineering，共有21個。研究證實新增的類別比舊的有較高的跨領域性，它們在多重指派比例(69% 比55%)、連結強度(6.36比5.11)、連結多樣性(4.4 比3.4)等指標上都比較高。
因為那些被分配到多個類別的期刊包含對不同學科有用的知識，即跨學科知識，本研究便是利用此一概念，證實了利用多重指派相關指標做為跨領域性的有用性，其結果的解釋需要加以留意，因為本研究相當仰賴ISI的主題分類，而此一分類系統並非完美。另外，本研究對於本身便包含多種跨學科性期刊的一般類別的效果也並比較沒有效果。
本研究能夠建立各學科的型態，區分出連結的類別主要為其他領域的大跨學科性(big interdisciplinarity)以及主要為本身領域的小跨學科性(small interdisciplinarity)，並且也證實新出現的學科主要為大跨學科性，具有較高的跨學科性。
The bibliometric methodology presented here provides a general overview of all scientific disciplines, with special attention to their interrelation.
Interdisciplinarity is measured through a series of indicators based on Institute for Scientific Information (ISI) multi-assignation of journals in subject categories.
Research areas and categories are described according to the quantity of their links (number of related categories) and their quality (with close or distant categories, diversity, and strength of links).
This differentiates “big” interdisciplinarity, which links distant categories, from “small” interdisciplinarity, in which close categories are related.
The most commonly accepted definitions come from the OECD (1998), in which multi-disciplinarity, interdisciplinarity, and transdisciplinarity are used to refer to increasing levels of interaction among disciplines. Thus, in multidisciplinary research, the subject under study is approached from different angles, using different disciplinary perspectives and integration is not accomplished. Interdisciplinary research leads to the creation of a theoretical, conceptual, and methodological identity, so more coherent and integrated results are obtained. Finally, transdisciplinarity goes one step further and it refers to a process in which convergence among disciplines is observed, and it is accompanied by a mutual integration of disciplinary epistemologies (Van den Besselaar & Heimer-
The most commonly accepted definitions come from the OECD (1998), in which multi-disciplinarity, interdisciplinarity, and transdisciplinarity are used to refer to increasing levels of interaction among disciplines. Thus, in multidisciplinary research, the subject under study is approached from different angles, using different disciplinary perspectives and integration is not accomplished. Interdisciplinary research leads to the creation of a theoretical, conceptual, and methodological identity, so more coherent and integrated results are obtained. Finally, transdisciplinarity goes one step further and it refers to a process in which convergence among disciplines is observed, and it is accompanied by a mutual integration of disciplinary epistemologies (Van den Besselaar & Heimeriks, 2001).
Among other useful methods, we can mention those that analyze the collaboration between authors or centers from different disciplines (Qiu, 1992; Qin et al., 1997; Bordons etal., 1999), or those focused on the distribution of references/citations over categories (Porter & Chubin, 1985; Urata,1990; Tomov & Mutafov, 1996; Bourke & Butler, 1998; Van Leeuwen & Tijssen, 2000; Van Raan & Van Leeuwen, 2002; Rinia et al., 2002).
In these studies, disciplines are frequently operationalized in terms of journal sets. Thus,“boundary crossing” authors have been identified as those who publish in journals from disciplines outside their own.
The analysis of cross-disciplinary citations in journal articles has been used for the study of knowledge exchange between disciplines in previous studies (Van Leeuwen &Tijssen, 2000; Rinia et al., 2002; NSF, 2002), in which articles were attributed to subfields on the basis of the classification of journals into categories.
We consider that the multi-assignation of journals to more than one category indicates the existence of cognitive links between disciplines, which can result in interdisciplinary research and whose evolution can be traced over time.
In this study, we assume that the different disciplines will be differently involved in cross-disciplinary activities, as has been previously stated (OECD, 1998), and we would like to test the sensitivity of several journal multi-assignation indicators to discriminate between disciplines and areas ac-cording to their degree of interdisciplinarity.
The classificatory scheme of knowledge we have used is the classification of journals into subject categories of the Institute for Scientific Information (ISI) in the Science Citation Index (SCI), Social Sciences Citation Index (SSCI), and the Arts & Humanities Citation Index (A&HCI). This classificatory scheme, which is updated periodically, in 1996 grouped approximately 8,000 journals into 224 categories. Each journal was assigned to at least one scientific category, and multi-assignation in up to five categories was frequently allowed.
In our study, interdisciplinarity in science was examined through the assignation of journals to more than one subject category. We assume that those journals that appear under more than one subject heading should be more interdisciplinary than those single-assigned.
From this starting point, a set of indicators, based on multi-assignation, was introduced to quantify and qualify categories according to their interdisciplinary links:
a) Percentage of multi-assigned journals per category.
b) Multi-assignation pattern. It distinguishes “internal links,” which are the result of the multi-assignation of journals to categories of the same area, and “external links,” created by the multi-assignation of journals to categories of different areas.
c) Diversity of relationships, calculated as the number of different links established by a given category, that is, the number of different categories that share journals.
d) Strength/intensity of the relationships established between two given categories. The number of links between two categories is normalized according to the size of each of the categories by means of the Salton index (Salton & McGill, 1983).
In Table 1, we can observe for each area the following data: number of journals included, average percentage of multi-assigned journals, and multi-assignation pattern. The percentage of multi-assigned journals ranged from 11% in Humanities to 56% in Biomedicine and Engineering. The most isolated area according to this indicator was Humanities.
The areas most externally related were Chemistry, Mathematics, and Physics, whereas Social Sciences, Humanities, and Engineering showed the highest internal multi-assignation rate.
The areas were grouped according to their multi-assignation percentage and pattern through hierarchical clustering analysis and, thus, four different groups were obtained (see clusters A–D in the last column of Table 1). It is interesting to note that Humanities and Social Sciences remain separated and are not included in any group of areas, as they show a different behavioral pattern.
Interdisciplinarity is measured through a series of indicators based on Institute for Scientific Information (ISI) multi-assignation of journals in subject categories.
Research areas and categories are described according to the quantity of their links (number of related categories) and their quality (with close or distant categories, diversity, and strength of links).
This differentiates “big” interdisciplinarity, which links distant categories, from “small” interdisciplinarity, in which close categories are related.
The most commonly accepted definitions come from the OECD (1998), in which multi-disciplinarity, interdisciplinarity, and transdisciplinarity are used to refer to increasing levels of interaction among disciplines. Thus, in multidisciplinary research, the subject under study is approached from different angles, using different disciplinary perspectives and integration is not accomplished. Interdisciplinary research leads to the creation of a theoretical, conceptual, and methodological identity, so more coherent and integrated results are obtained. Finally, transdisciplinarity goes one step further and it refers to a process in which convergence among disciplines is observed, and it is accompanied by a mutual integration of disciplinary epistemologies (Van den Besselaar & Heimer-
The most commonly accepted definitions come from the OECD (1998), in which multi-disciplinarity, interdisciplinarity, and transdisciplinarity are used to refer to increasing levels of interaction among disciplines. Thus, in multidisciplinary research, the subject under study is approached from different angles, using different disciplinary perspectives and integration is not accomplished. Interdisciplinary research leads to the creation of a theoretical, conceptual, and methodological identity, so more coherent and integrated results are obtained. Finally, transdisciplinarity goes one step further and it refers to a process in which convergence among disciplines is observed, and it is accompanied by a mutual integration of disciplinary epistemologies (Van den Besselaar & Heimeriks, 2001).
Among other useful methods, we can mention those that analyze the collaboration between authors or centers from different disciplines (Qiu, 1992; Qin et al., 1997; Bordons etal., 1999), or those focused on the distribution of references/citations over categories (Porter & Chubin, 1985; Urata,1990; Tomov & Mutafov, 1996; Bourke & Butler, 1998; Van Leeuwen & Tijssen, 2000; Van Raan & Van Leeuwen, 2002; Rinia et al., 2002).
In these studies, disciplines are frequently operationalized in terms of journal sets. Thus,“boundary crossing” authors have been identified as those who publish in journals from disciplines outside their own.
The analysis of cross-disciplinary citations in journal articles has been used for the study of knowledge exchange between disciplines in previous studies (Van Leeuwen &Tijssen, 2000; Rinia et al., 2002; NSF, 2002), in which articles were attributed to subfields on the basis of the classification of journals into categories.
We consider that the multi-assignation of journals to more than one category indicates the existence of cognitive links between disciplines, which can result in interdisciplinary research and whose evolution can be traced over time.
In this study, we assume that the different disciplines will be differently involved in cross-disciplinary activities, as has been previously stated (OECD, 1998), and we would like to test the sensitivity of several journal multi-assignation indicators to discriminate between disciplines and areas ac-cording to their degree of interdisciplinarity.
The classificatory scheme of knowledge we have used is the classification of journals into subject categories of the Institute for Scientific Information (ISI) in the Science Citation Index (SCI), Social Sciences Citation Index (SSCI), and the Arts & Humanities Citation Index (A&HCI). This classificatory scheme, which is updated periodically, in 1996 grouped approximately 8,000 journals into 224 categories. Each journal was assigned to at least one scientific category, and multi-assignation in up to five categories was frequently allowed.
In our study, interdisciplinarity in science was examined through the assignation of journals to more than one subject category. We assume that those journals that appear under more than one subject heading should be more interdisciplinary than those single-assigned.
From this starting point, a set of indicators, based on multi-assignation, was introduced to quantify and qualify categories according to their interdisciplinary links:
a) Percentage of multi-assigned journals per category.
b) Multi-assignation pattern. It distinguishes “internal links,” which are the result of the multi-assignation of journals to categories of the same area, and “external links,” created by the multi-assignation of journals to categories of different areas.
c) Diversity of relationships, calculated as the number of different links established by a given category, that is, the number of different categories that share journals.
d) Strength/intensity of the relationships established between two given categories. The number of links between two categories is normalized according to the size of each of the categories by means of the Salton index (Salton & McGill, 1983).
In Table 1, we can observe for each area the following data: number of journals included, average percentage of multi-assigned journals, and multi-assignation pattern. The percentage of multi-assigned journals ranged from 11% in Humanities to 56% in Biomedicine and Engineering. The most isolated area according to this indicator was Humanities.
The areas most externally related were Chemistry, Mathematics, and Physics, whereas Social Sciences, Humanities, and Engineering showed the highest internal multi-assignation rate.
The areas were grouped according to their multi-assignation percentage and pattern through hierarchical clustering analysis and, thus, four different groups were obtained (see clusters A–D in the last column of Table 1). It is interesting to note that Humanities and Social Sciences remain separated and are not included in any group of areas, as they show a different behavioral pattern.
To quantify the diversity of relationships, the number of different links established between pairs of categories was calculated.
The strength of links between categories was analyzed through the Salton index. The higher the Salton index value between two given categories, the greater is their relationship or strength of links.
The research areas and the categories can be described through the combination of the different indicators introduced: multi-assignation percentage, external multi-assignation rate, diversity, and strength of links.
Areas were grouped according to the multi-assignation percentage, the percentage of external links, and the strength of links. The variable “diversity of links” was removed due to its low discriminating power found in the analysis.
The fact that a mainly basic area such as Biomedicine and the prototype of applied area that is Engineering/Technology are grouped together is surprising, and it indicates again that interdisciplinarity is not related to whether research is of the applied or basic type.
Those journals that are assigned to more than one category are to be read by different communities of scientists, so they must presumably include knowledge useful for different disciplines, that is, interdisciplinary knowledge.
The strength of links between categories was analyzed through the Salton index. The higher the Salton index value between two given categories, the greater is their relationship or strength of links.
The research areas and the categories can be described through the combination of the different indicators introduced: multi-assignation percentage, external multi-assignation rate, diversity, and strength of links.
Areas were grouped according to the multi-assignation percentage, the percentage of external links, and the strength of links. The variable “diversity of links” was removed due to its low discriminating power found in the analysis.
The fact that a mainly basic area such as Biomedicine and the prototype of applied area that is Engineering/Technology are grouped together is surprising, and it indicates again that interdisciplinarity is not related to whether research is of the applied or basic type.
Those journals that are assigned to more than one category are to be read by different communities of scientists, so they must presumably include knowledge useful for different disciplines, that is, interdisciplinary knowledge.
Interdisciplinary indicators based on ISI multi-assignation of journals to categories has proved useful in providing a deeper understanding of the relations between disciplines. However, the results should be analyzed with caution since they are highly dependent on the ISI classification scheme, which is not perfect.
This approach allows the establishment of a typology of disciplines, which differentiates those responsible for the “big interdisciplinarity” (predominance of distant links, that is, those between different areas) and the “small interdisciplinarity” (links between close disciplines or disciplines of the same area).
New emerging disciplines are highly interdisciplinary, and show a predominance of the “big interdisciplinarity.”
This approach allows the establishment of a typology of disciplines, which differentiates those responsible for the “big interdisciplinarity” (predominance of distant links, that is, those between different areas) and the “small interdisciplinarity” (links between close disciplines or disciplines of the same area).
New emerging disciplines are highly interdisciplinary, and show a predominance of the “big interdisciplinarity.”
In these studies, disciplines are
frequently operationalized in terms of journal sets. Thus,
“boundary crossing” authors have been identified as those
who publish in journals from disciplines outside their own.
In these studies, disciplines are
frequently operationalized in terms of journal sets. Thus,
“boundary crossing” authors have been identified as those
who publish in journals from disciplines outside their own.
iks, 2001).
The most commonly accepted
definitions come from the OECD (1998), in which multi-
disciplinarity, interdisciplinarity, and transdisciplinarity are
used to refer to increasing levels of interaction among
disciplines. Thus, in multidisciplinary research, the subject
under study is approached from different angles, using
different disciplinary perspectives and integration is not
accomplished. Interdisciplinary research leads to the cre-
ation of a theoretical, conceptual, and methodological iden-
tity, so more coherent and integrated results are obtained.
Finally, transdisciplinarity goes one step further and it refers
to a process in which convergence among disciplines is
observed, and it is accompanied by a mutual integration of
disciplinary epistemologies (Van den Besselaar & Heimer-
iks, 2001).
