本研究利用作者共被引分析(author cocitation analysis)分析STS領域的社會結構，探討做為次領域間連結的作者或研究機構。本研究將STS領域分為STS的量化研究次領域(the qualitative STS sub-field)、STS的質性研究次領域(the qualitative STS sub-field)和政策導向次領域(the policy oriented sub-field)，並且以Scientometrics期刊為STS的量化研究的代表，Social Studies of Science和Science, Technology and Human Values兩種期刊代表STS的質性研究，Research Policy則是STS政策研究的代表。針對1986到1997年間在這些期刊上被引用超過25次的229位作者，建立他們的共被引矩陣，然後進行因素分析(factor analysis)，查看這些作者被歸類的情形，並且與上述的次領域進行比較分析。此外，本研究也探討被不同次領域引用的作者、不同次領域之間的作者的合作關係以及有多少位作者在不同的次領域發表論文？
Table 1表示762、305、304及569位作者分別曾在Scientometrics、Social Studies of Science、Science, Technology and Human Values以及Research Policy等期刊發表論文，Scientometrics和Research Policy的作者平均在對應的期刊上發表1.5及1.7篇，比Social Studies of Science和Science, Technology and Human Values的1.1篇來得高。曾在四種期刊發表論文的作者則是1756位，平均每位作者發表的論文數為1.4。
共有759個機構曾在四種期刊上發表論文，但只有少數的機構有較高的生產力，例如超過11篇論文的機構僅有41個。此外，從Table 2也可以發現有些高生產力機構的論文是在不同次領域的期刊上發表。
共有65個國家在四種期刊上發表論文，其中大多數的國家(57個)有在Scientometrics上發表，但其他三種期刊都僅有約半數的國家有發表的紀錄。
將作者共被引矩陣進行因素分析後，較大的因素共有7個，依作者撰寫論文的內容將各因素命名。其中第1個因素和第6個因素間有很大的關係，第1個因素有大半數的作者的次高負荷是落在第6個因素上，反之亦然，第1個因素和第6個因素的研究主題為科技政策相關的STS研究。第2個因素的研究主題為STS的質性研究。第3個因素和第4個因素、第5個因素以及第7個因素彼此間的作者有關係，這些因素可以視為是STS的量化研究，進一步來說，第4個因素、第5個因素和第7個因素的主題分別是科學社會學(Sociology of Science)、詞語共現分析和資訊計量學。
接下來，Table 4 分析各次領域的專家(specialists)以及兼通兩門或以上的通才(generalists)。本研究將專家定義為在該次領域發表的論文數超過該領域論文總數0.69%以上的作者，STS的量化研究、質性研究和政策導向研究等次領域各有31、23和41位。量化研究次領域的專家並且也發表質性研究相關論文的作者有6位，反之質性研究次領域的專家並且也發表量化研究相關論文的作者只有2位。量化研究次領域的專家同時發表政策導向相關論文的作者有14位，政策導向研究次領域的專家並且也發表量化研究相關論文的作者則有11位。從以上數據顯示，量化研究與其他兩個次領域的關係主要是由量化研究次領域的研究者在維繫著，也就是量化研究次領域的研究者是主要的跨邊界者(boundary spanners)。
Table 4上也可以發現一些從質性研究次領域跨越政策導向研究的研究者，這個研究結果修正了先前認為質性研究次領域比較獨立的看法。
The differentiation of scientific fields into sub-fields can be studied on the level of the ‘scientific content’ of the sub-field, that is on the level of the products, as well as on the level of the ‘social structures’ of the sub-field, that is on the level of the producers of the content.
By comparing the behavior of the constructs with the behavior of the constructors, we are able to demonstrate the analytical distinction between a cognitive and a social approach in an empirical way.
Although we are able to distinguish analytically between the cognitive and social dimension of the development of the research field, we find similar patterns of differentiation on the social level too. At the same time, this differentiation differs in some respects from the cognitive differentiation pattern.
Consequently, the social and the cognitive dimensions of the STS field are not independent – as no serious STS scholar would argue – but also not identical, as radical constructivists claim, but are strongly interacting.
It was claimed that scientometrics has to focus more on the role it can play for qualitative STS, and that scientometric researchers should refrain from sterile data and mathematics. It was felt that scientometric results have to be carefully interpreted from a substantial perspective, to be meaningful for S&T policy.
There, we showed that the journals Social Studies of Science (SSS) and Science, Technology and Human Values (STHV) form a reasonable operationalization of the qualitative STS sub-field. Research Policy represents the policy oriented sub-field, and Scientometrics can be used as a representation of the quantitative STS sub-field. These journals are central in STS as they have the highest impact factors in their respective sub-fields.
In this paper we will use the same boundary of STS to analyze the social structure of the field: who are the authors and what are the research groups in the field as defined by the mentioned journals? Do they function as the ties between the various sub-fields?
Data about authors and institutional affiliation can be found on the CD-ROM version of the Social Science Citation Index (SSCI). We downloaded the full records for all publications in the four journals for the period 1986-1997.* This resulted in a database with 3579 records. ... Finally, as is usual in scientometric studies, for further analysis we restricted the database to Articles, Reviews, Notes, and Letters, and excluded other document types. This resulted in a final set of 1787 documents.
Referring to a text may indicate the use of a knowledge claim to support one’s own position, or to oppose to. Referring to persons, on the other hand, may indicate the existence of a social relationship. Therefore we will use author co-citation analysis as a first methodology to analyze the social structure of the STS field. In this way, we will describe the STS field in terms of clusters of authors that are placed near each other by the scholars active in the field.
Using the prepared database and bibexcel, an author co-citation matrix has been produced of all cited 229 authors with more than 25 citations over the 1986-1997 period. Factor-analyzing (principal component analysis, varimax rotation with Kaiser normalization) this matrix results in clusters of authors, and the question is whether these clusters differ from the three sub-fields of qualitative, quantitative, and policy oriented STS.
If a communication system shows considerable segregation, individual researchers (or institutes) could play the role as weak ties [3] between the sub-fields.
(i) Authors can refer to materials from other sub-fields. We classify these authors as being active on the borders of the sub-fields. The border between sub-fields A and B is then defined as the authors of papers in sub-field A referring to papers in sub-field B, and the other way around. How densely populated are the borders between the subfields?
(ii) Authors can cooperate with colleagues active in the other sub-fields. Even if authors specialize, research groups and institutions may cover more sub-fields, and this could indicate social integration of the field on a more informal level of communication.
(iii) Generalist authors work in various sub-fields. Do many authors publish in more than one sub-field, or do we see a specialization and differentiation on the level of individual scholars? How many generalists can be found among researchers and institutions? The larger numbers we find, the stronger is the degree of communication between the sub-fields.
The average number of authors per article is 1.4, but this figure is higher in Scientometrics (1.5) and in Research Policy (1.7), but considerable lower (1.1) in the two qualitative STS journals.
As expected, the number of frequently publishing institutes is rather small, compared to the grand total.
If we aggregate one more step, to the level of countries, we find 65 countries active in the STS field, of which some 57 are active within scientometrics. However, only half of the countries are publishing in the qualitative journals SSS and STHV. The same is true for Research Policy.
Factor analyzing the author co-citation matrix resulted in a solution of 22 factors with an eigenvalue larger than 1. Inspecting the scree plot shows that seven factors dominate the structure, and these factors explain more than 70% of the total variance. More than 90% of the 220 cited authors have their highest factor score on one of these seven factors.
The authors in Factor 1 are within science & technology policy studies and in research & innovation management studies, or in related fields in management and economics. The same holds for the small Factor 6. Half of the authors in Factor 1 have a relatively high second factor loading in Factor 6, and all authors with their highest loading on factor 6 do have a high second loading on Factor 1.
Authors with their highest factor loading on Factor 2 all belong to qualitative STS, and they generally do not load on other factors.
Factor 3 represents quantitative STS. Most authors with the highest loading on Factor 4 can be characterized as traditional sociology of science (e.g., Merton). Factor 5 represents coword analysis, and Factor 7 represents informetrics and scientometric distributions (e.g., Bradford and Lotka). Between the Factors 3, 4, 5, and 7 we find a considerable ‘interfactorial complexity’: the authors loading highest on Factor 3 often have a substantial second loading on one of the Factors 4, 5, or 7. The same is true the other way around.
Therefore I also created the author co-citation matrix of all authors with more than 25 citations over the whole period with the highest loading on the Factors 3, 4, 5, or 7. Authors that have a second loading on these factors of more than 0.2 are also included. This set of authors represents the sub-field scientometrics.
Factor-analyzing this matrix in a similar way results in seven substantial factors. Inspection of the factors shows that they represent the following research foci: Policy oriented scientometrics (Factor 1); Empirical science & technology studies (Factor 2); Coword analysis (Factor 3); Scientometric distributions (Factor 4); Critique of scientometrics (Factor 5); Patent studies (Factor 6); Economics of technical change (Factor 7). This result corroborates that the method is suited for analyzing the fine structure of research fields.
If we now summarize these findings, the factor-structure of the co-citation matrix of STS reproduces the clear split between policy oriented STS (Factor 1 plus 6), qualitative STS (Factor 2), and quantitative STS (Factors 3, 4, 5, 7), while at the same time showing some internal differentiation in the sub-field of scientometrics. In other words, the author co-citation analysis reveals a similar structure as the journal-journal citation analysis did.8
Firstly, we distinguish the groups of specialists, which consist of the authors with relatively high numbers of publications in one of the various sub-fields of STS. We consider a scholar as specialist in one of the sub-fields, if he or she is (co-) author of at least 6, 4, or 3 publications respectively in quantitative, qualitative, or policy oriented STS. In this way, the threshold is about the same in the three sub-fields: 0.77%, 0.69%, and 0.73%.
Secondly, we have the semi-generalists, the groups of authors active in two of the three sub-fields each. A semi-generalist is defined as an author who has published at least two publications in two of the three sub-fields.
Finally we have the group of generalists, publishing in all the three sub-fields, again based on at least two publications per sub-field.
The number of specialists in Scientometrics is 31, and only six of them have published in SSS or STHV. The other way around we identified only 2 authors. This implies that the more quantitative researchers maintain the relations between these two sub-fields
The number of Scientometrics authors also publishing in Research Policy is much higher, and some 45% of the scientometrics specialists also work – at least incidentally – on S&T policy topics. Researchers frequently publishing in Research Policy publish a little less (27%) in Scientometrics, but this is still a substantial number.
This underlines our earlier conclusion that research policy and management is related to scientometrics for the part of using scientometrics in research evaluation, but not much wider.8
Between Scientometrics and Research Policy, as well as between Scientometrics and SSS/STHV, most of the authors who maintain the relation have most publications in Scientometrics, and generally only a single publication in one of the other journals. This implies that the relations between the sub-fields (also the very weak one’s) are maintained to a large extent by scientometricians.
Between Research Policy and SSS/STHV the picture is more balanced, with a weak emphasis on the SSS/STHV authors. The number of authors publishing both in qualitative STS and S&T policy studies is very low, as is the number of authors publishing both in quantitative STS and in qualitative STS. Only the number of authors publishing in both quantitative STS and S&T policy studies is substantial.
Lowering the threshold increases the number of (semi-)generalists, but of course most of them have a very low number of publications, and the scientometricians are the boundary spanners, much more than the others.
However, a larger number of qualitative authors than expected is also active in the S&T policy studies. Only this latter finding modifies slightly our earlier conclusion that qualitative STS is an isolated sub-field.
We use a 3% threshold, and various organizations that exceed this threshold in one of the sub-fields are in Table 6. Three of the eight organizations are specialized in only one sub-field. Four others are specialized in two sub-fields, and only one organization is a generalist one, and active in three sub-fields.
In other words, there is a relatively low level of specialization here, as most of the institutions seem to be rather active in more sub-fields.
If we decrease the threshold to 2%, another 15 institutions count as specialists. However, of these 15 institutions only a few are active in more sub-fields. This implies that the most productive institutions within STS are also the broadest in their covering of the field.
Where the cognitive analysis showed that the relationship between scientometrics and S&T policy studies is stronger than the relations between qualitative and quantitative STS,8 on the level of the conferences (and as shown before, on the level of research institutes) it is the other way around. In other words, the institutional structures and the cognitive structures are not identical.
If we summarize the findings, we see that the cognitive patterns of integration and (mainly) differentiation to a large extent are visible within the social structure of the field.
The social relations between quantitative STS and policy oriented STS are similar to the cognitive relations between the two sub-fields. The links, however, between the two sub-fields are only between a substantial part of scientometrics and a small part of S&T policy studies, namely the part focusing on evaluation and performance studies.
The larger part of S&T policy studies is on technological innovation and on evolutionary approaches to technical change, and these research topics are not related to the research front in scientometrics, as the author co-citation analysis underlines.
Most importantly, we found that the interaction between qualitative and policy oriented STS is much stronger on the social level of authors and institutions than on the cognitive level of documents.
This may explain why the discussants in the panel session quoted earlier in this paper saw different divides, than the one’s I revealed in Ref. 8: the social structure of the STS field is not identical to its cognitive structure.
Within the mainstream of STS it is generally accepted that the production of knowledge and the grounding of knowledge claims consists of a ‘seamless web’ of cognitive and social elements.
(ii) Authors can cooperate with colleagues active in the other sub-fields. Even if authors specialize, research groups and institutions may cover more sub-fields, and this could indicate social integration of the field on a more informal level of communication.
(iii) Generalist authors work in various sub-fields. Do many authors publish in more than one sub-field, or do we see a specialization and differentiation on the level of individual scholars? How many generalists can be found among researchers and institutions? The larger numbers we find, the stronger is the degree of communication between the sub-fields.
The average number of authors per article is 1.4, but this figure is higher in Scientometrics (1.5) and in Research Policy (1.7), but considerable lower (1.1) in the two qualitative STS journals.
As expected, the number of frequently publishing institutes is rather small, compared to the grand total.
If we aggregate one more step, to the level of countries, we find 65 countries active in the STS field, of which some 57 are active within scientometrics. However, only half of the countries are publishing in the qualitative journals SSS and STHV. The same is true for Research Policy.
Factor analyzing the author co-citation matrix resulted in a solution of 22 factors with an eigenvalue larger than 1. Inspecting the scree plot shows that seven factors dominate the structure, and these factors explain more than 70% of the total variance. More than 90% of the 220 cited authors have their highest factor score on one of these seven factors.
The authors in Factor 1 are within science & technology policy studies and in research & innovation management studies, or in related fields in management and economics. The same holds for the small Factor 6. Half of the authors in Factor 1 have a relatively high second factor loading in Factor 6, and all authors with their highest loading on factor 6 do have a high second loading on Factor 1.
Authors with their highest factor loading on Factor 2 all belong to qualitative STS, and they generally do not load on other factors.
Factor 3 represents quantitative STS. Most authors with the highest loading on Factor 4 can be characterized as traditional sociology of science (e.g., Merton). Factor 5 represents coword analysis, and Factor 7 represents informetrics and scientometric distributions (e.g., Bradford and Lotka). Between the Factors 3, 4, 5, and 7 we find a considerable ‘interfactorial complexity’: the authors loading highest on Factor 3 often have a substantial second loading on one of the Factors 4, 5, or 7. The same is true the other way around.
Therefore I also created the author co-citation matrix of all authors with more than 25 citations over the whole period with the highest loading on the Factors 3, 4, 5, or 7. Authors that have a second loading on these factors of more than 0.2 are also included. This set of authors represents the sub-field scientometrics.
Factor-analyzing this matrix in a similar way results in seven substantial factors. Inspection of the factors shows that they represent the following research foci: Policy oriented scientometrics (Factor 1); Empirical science & technology studies (Factor 2); Coword analysis (Factor 3); Scientometric distributions (Factor 4); Critique of scientometrics (Factor 5); Patent studies (Factor 6); Economics of technical change (Factor 7). This result corroborates that the method is suited for analyzing the fine structure of research fields.
If we now summarize these findings, the factor-structure of the co-citation matrix of STS reproduces the clear split between policy oriented STS (Factor 1 plus 6), qualitative STS (Factor 2), and quantitative STS (Factors 3, 4, 5, 7), while at the same time showing some internal differentiation in the sub-field of scientometrics. In other words, the author co-citation analysis reveals a similar structure as the journal-journal citation analysis did.8
Firstly, we distinguish the groups of specialists, which consist of the authors with relatively high numbers of publications in one of the various sub-fields of STS. We consider a scholar as specialist in one of the sub-fields, if he or she is (co-) author of at least 6, 4, or 3 publications respectively in quantitative, qualitative, or policy oriented STS. In this way, the threshold is about the same in the three sub-fields: 0.77%, 0.69%, and 0.73%.
Secondly, we have the semi-generalists, the groups of authors active in two of the three sub-fields each. A semi-generalist is defined as an author who has published at least two publications in two of the three sub-fields.
Finally we have the group of generalists, publishing in all the three sub-fields, again based on at least two publications per sub-field.
The number of specialists in Scientometrics is 31, and only six of them have published in SSS or STHV. The other way around we identified only 2 authors. This implies that the more quantitative researchers maintain the relations between these two sub-fields
The number of Scientometrics authors also publishing in Research Policy is much higher, and some 45% of the scientometrics specialists also work – at least incidentally – on S&T policy topics. Researchers frequently publishing in Research Policy publish a little less (27%) in Scientometrics, but this is still a substantial number.
This underlines our earlier conclusion that research policy and management is related to scientometrics for the part of using scientometrics in research evaluation, but not much wider.8
Between Scientometrics and Research Policy, as well as between Scientometrics and SSS/STHV, most of the authors who maintain the relation have most publications in Scientometrics, and generally only a single publication in one of the other journals. This implies that the relations between the sub-fields (also the very weak one’s) are maintained to a large extent by scientometricians.
Between Research Policy and SSS/STHV the picture is more balanced, with a weak emphasis on the SSS/STHV authors. The number of authors publishing both in qualitative STS and S&T policy studies is very low, as is the number of authors publishing both in quantitative STS and in qualitative STS. Only the number of authors publishing in both quantitative STS and S&T policy studies is substantial.
Lowering the threshold increases the number of (semi-)generalists, but of course most of them have a very low number of publications, and the scientometricians are the boundary spanners, much more than the others.
However, a larger number of qualitative authors than expected is also active in the S&T policy studies. Only this latter finding modifies slightly our earlier conclusion that qualitative STS is an isolated sub-field.
We use a 3% threshold, and various organizations that exceed this threshold in one of the sub-fields are in Table 6. Three of the eight organizations are specialized in only one sub-field. Four others are specialized in two sub-fields, and only one organization is a generalist one, and active in three sub-fields.
In other words, there is a relatively low level of specialization here, as most of the institutions seem to be rather active in more sub-fields.
If we decrease the threshold to 2%, another 15 institutions count as specialists. However, of these 15 institutions only a few are active in more sub-fields. This implies that the most productive institutions within STS are also the broadest in their covering of the field.
Where the cognitive analysis showed that the relationship between scientometrics and S&T policy studies is stronger than the relations between qualitative and quantitative STS,8 on the level of the conferences (and as shown before, on the level of research institutes) it is the other way around. In other words, the institutional structures and the cognitive structures are not identical.
If we summarize the findings, we see that the cognitive patterns of integration and (mainly) differentiation to a large extent are visible within the social structure of the field.
The social relations between quantitative STS and policy oriented STS are similar to the cognitive relations between the two sub-fields. The links, however, between the two sub-fields are only between a substantial part of scientometrics and a small part of S&T policy studies, namely the part focusing on evaluation and performance studies.
The larger part of S&T policy studies is on technological innovation and on evolutionary approaches to technical change, and these research topics are not related to the research front in scientometrics, as the author co-citation analysis underlines.
Most importantly, we found that the interaction between qualitative and policy oriented STS is much stronger on the social level of authors and institutions than on the cognitive level of documents.
This may explain why the discussants in the panel session quoted earlier in this paper saw different divides, than the one’s I revealed in Ref. 8: the social structure of the STS field is not identical to its cognitive structure.
Within the mainstream of STS it is generally accepted that the production of knowledge and the grounding of knowledge claims consists of a ‘seamless web’ of cognitive and social elements.
沒有留言:
張貼留言