Gretchen Whitney

Abstract:

South American authorship patterns in 16 major scientific and technical bibliographic databases of the DIALOG Information Services were examined, for the period 1970-1990. Where trends could be identified, it was found that there was an increasing number of records with South American authors during the 1970's, then a decline in the 1980's. Some steady increases of South American authorship were found in BIOSIS, INSPEC, and MEDLINE. The most heavily represented countries in the 16 databases studied were Brazil, Argentina, Venezuela, and Chile.

(Support for this research was provided by a research grant from DIALOG Information Services, for which the author is most grateful. The author is also grateful for the assistance of Sarah Dorsey in the preparation of data and charts for analysis.)

 Introduction

The visibility of a country’s publications in the international scientific and technical literature is an important measure of its participation in the international scientific arena, and is intricately entangled in issues of national pride, cultural identity, and even a sense of ‘fair play’. While other measures may be used to assess scientific development in a given country — such as governmental support of research, numbers of graduate students in various disciplines, and travel patterns of scientists — bibliometric studies of scientific literatures remain a relatively convenient and readily understood resource.

These measures are important to libraries in the United States, in assessing the nature and amount of literature available from various regions of the world in these files. For Third World libraries, they are important indicators of the degree to which their country’s literature is reaching an international audience. While databases primarily serve as reference tools, they also represent or reflect collection development policies and practices in the separate provider organizations.

This research focuses on South American authorship in major scientific and technical databases available for examination on the database system of DIALOG Information Services (Palo Alto, California), during 1970–1990. This work complements a similar study on Asian authorship patterns [1] and work in progress assessing European and African authorship trends. Two earlier works also provide details concerning the overall research program, which involves the investigation of issues in the internationality of scientific and technical bibliographic databases [2], [3].

One of the basic assumptions of these studies is that increasing figures have a positive social or cultural value, in particular when the overall trends for the number of records included in a database also are increasing. There has been concern in the political arena about the ability of the scientist in the United States to identify and acquire knowledge of foreign science and technology, particularly research from Japan, China, and (recently to a lesser degree) the Soviet Union. Rising figures, therefore, suggest some increase in access, at least quantitatively.

These studies also assume that, except in special cases [4] there should be a simple increase in the number of records added from a given country, no matter how slight, in these disciplines and from these countries. The majority of databases report addition of an increasing number of records each year. There is no good reason to assume the contrary position: that countries are decreasing production of scientific and technical work. Looking for patterns in DIALOG databases does disregard the fact that there are local and regional databases being developed that identify and present South American research. Examination of those databases is beyond the scope of the current work. Further, these files are less readily available to American scientists than those on DIALOG and similar vendor systems.

A glance at recent issues of Scientometrics will show that bibliometric studies assessing the scientific and technical output of countries have frequently been undertaken using the Institute for Scientific Information's Science Citation Index. Indeed, using a database produced by a single provider clarifies policy issues and changes more readily. In contrast, use of the DIALOG databases provides files and coding practices developed by organizations which are field specialists: the American Chemical Society for CA Search, the National Library of Medicine for MEDLINE, etc. Once the groundwork has been accomplished by laying out the patterns for DIALOG, comparative work can be undertaken.

 Methodology

The major bibliographic databases in the DIALOG system, containing records for documents published in approximately 1970–1990, and aiming to cover the world’s literature in their disciplines, have been identified. They are Aerospace Abstracts, Agricola, BIOSIS, Chemical Abstracts, COMPENDEX, GeoRef, INSPEC, ISMEC, Life Sciences, Math Sciences, MEDLINE, NTIS, Oceanic Abstracts, PASCAL, and Psyclnfo. Sociological Abstracts, the National Criminal Justice Reference Service, and the Public Affairs Information Service were studied in earlier papers, and found to be less fruitful for his type of research. Countries of the world were divided into the geographic regions of Asia, Africa, South America, Europe and North America, according to the classification used by the United Nations in the Unesco Statistical Yearbook (1980 and 1987). This paper concerns South American authorship, and the specific countries of concern are shown in the Appendices and Tables. For each database, all country names were searched to determine those countries that accounted for the largest number of records represented in the database for the South American region. Further, those countries were identified which account for 90% of all records for each database for the region, or for which there were at least 100 records in the database. South American authors are not strongly represented in the databases; even a contribution of fewer than a thousand records may illustrate a pattern that is the result of little more than random chance.

Records in the databases contain a specific field indicating the institutional and national affiliation of the authors of a given paper. The South American country names were searched in each of the databases, for records by year of publication for the years 1970–1990. Searching was conducted in the Spring of 1991.

 Results

Diversity of Databases
One measure of the diversity of a database is the proportion of the database contributed by the most productive countries. If, for example, just three countries make up 95% of the South American contribution to a database, it can be said to be less diverse than a database requiring 20 countries to account for 95% of its South American records.

The measures for the databases studied are in Table 1. They are listed in decreasing value, the lower number indicating that fewer countries proportionately contributed 95% of the South American portion of the database.

This measure makes the most sense at the upper and lower ends of the scale. The table suggests that relatively few countries form the bulk of South American representation in Chemical Abstracts, INSPEC, Aerospace Abstracts, and PASCAL; while more countries are needed for COMPENDEX, Agricola, Oceanic Abstracts, and BIOSIS. MATHSCI is an anomaly, because it includes materials from so few South American countries to begin with. These results are in contrast with European patterns of diversity. Note that the majority of the files fall in the extreme ranges (above 55, below 45). For Europe, the majority fall in the mid–range, hovering around .50. For Africa, the patterns are similar to South America, in that the figures fall at the extremes.

Overall Representation in Databases
Table 2 shows the total number of records for each country in all of the databases. Brazil clearly dominates the region for these files. Only Brazil, Argentina, Chile, Venezuela, and Colombia were represented by over 1,000 records in any one database.

Brazil’s largest output was represented in Chemical Abstracts, where it is comparable to Yugoslavia, with 32,000 records. For perspective, it may be noted that France produced ten times that number in the same period. The next highest coverage for Brazil was in PASCAL, with 18,000 records, comparable to Finland or Austria in Europe. In Aerospace Abstracts, Brazil, Argentina and Chile are the largest contributors, Brazil the strongest with 5,300 records, the others trailing with 1,400 records. In BIOSIS, the same three countries dominate, but in a more even distribution: Argentina has 8,100 records, Brazil has 7,400, and Chile has 5,700.

The contributions from Argentina to Chemical Abstracts are out of character in the emerging profile of South America. Only 126 records were found, as opposed to 32,761 for Brazil and 5,764 for Chile. This may reflect a problem with the database. BIOSIS, on the other hand, reported 8,124 records from Argentina and 7,474 from Brazil. (Note that this is one of two cases in which Argentina leads Brazil;, the other by a slim margin of 2,761 to 2,168 in MEDLINE).

Chile is behind Brazil and Argentina in most cases. The exceptions are in Chemical Abstracts, as noted above, and in COMPENDEX. In COMPENDEX, Chile is the largest contributor, with 941 records, followed by Brazil with 234. Chile approaches Brazil in Oceanic Abstracts, with 431 and 458 records respectively. In no other cases does Brazil show lower numbers than any other country.

Venezuela contributed its largest number of records to PASCAL and BIOSIS; Colombia’s largest numbers are in Chemical Abstracts and PASCAL.

In examining individual country production in databases, one might expect an emphasis in South America on medicine and agriculture, over disciplines such as aerospace science. If there is such an emphasis, it is not reflected in these databases. MEDLINE and Agricola each have about 10,000 records (from all South America). Aerospace Abstracts is not far behind, at nearly 9,000 records from all countries, and that is the lowest number for all databases.

Ranking of Countries in Databases
Appendix 1 presents the countries as they rank in the various databases. The countries presented are those which made up 95% of each database, no matter how many records are included for each country. Production from South America is very sparsely represented: there are only 50 instances of more than a thousand records in a given year, 15 in PASCAL alone, out of a possible 250 cases. The vast majority of countries have fewer than a hundred records for the 20-year period.

Trends in Authorship
Appendix 2 presents trends for these countries in the various databases. It is based on Appendix 1, in that the relative rankings have been preserved. The symbols identify those countries in each database that increased or peaked in contributions over the period of study (primarily 1970–1989, although in BIOSIS it is 1980–1989, and MATHSCI and MEDLINE it is 1985–89, because the databases did not code consistently until then). Appendix 2 gives an abbreviated form of the data, so that trends can be assessed at a glance. Clearly, the cases labeled INC represent different levels of increase. The actual numbers are provided in Appendix 3.

All in all, 70 ‘cases’ were identified, of all types. A case is an account of the number of records, by publication year, from one country in a given database. The countries were identified by the criteria noted above. Several examples of cases appear below. Cases were classed into four groups: (a) increasing; (b) increasing until a given year in the 1980s, then declining for the remainder of the period; (c) decreasing; (d) stable or erratic, showing no definite trend; and (e) too small to classify, with fewer than 100 records over the 20–year period. The 100 record cut–off is generous: it is barely justifiable that trends for even 1,000 records would be the result of any thing other than random chance. ‘Increasing’ and ‘decreasing’ are tendencies shown by the number of records for each year of publication. Contrary to earlier studies of database attributes, none of these cases could be considered in decline as an overall trend. Very few cases were definite trends of increase, however.

Examining the cases removed because of their small numbers also identifies the databases which have virtually been removed from consideration. All of the cases for Oceanic Abstracts, ISMEC, and nearly all of PsycINFO, NTIS, and COMPENDEX were removed. Half of those for MATHSCI and Agricola were removed.

The numbers of records falling into these trend categories are shown in Table 3.

Trends and Comparisons with Database Overall Patterns
One might hypothesize that the patterns reflected by a single country in a database might reflect the pattern overall in a database: that as the overall amount of literature that was added to the file grew each year, an equivalent increase would be seen in any given country. In terms of overall numbers of records, BIOSIS, Chemical Abstracts, INSPEC, MATHSCI, MEDLINE all show definite increases each year in the number of records added. PsycINFO, PASCAL, Oceanic Abstracts, NTIS, and GeoRef, all show increases until 1987, when they begin to decline. COMPENDEX increases until 1984, when decline sets in. Life Sciences is erratic. Agricola increases its numbers between 1977 and 1979, is stable for two years, and then begins a steady decline from 150,000 records to just under 40,000 records added each year. ISMEC is stable from 1974–1981 and then declines.

This section identifies trends for individual countries. They are also presented in Appendix 2. Those cases roughly matching the overall pattern of the database are asterisked (*). Seven cases maintained the increase through 1989, and can be considered increasing. These are BIOSIS documents with authors from Brazil (*) and Peru (*); INSPEC for Brazil (*) and Chile (*); and MEDLINE documents with authors from Chile (*), Brazil (*), and Argentina (*). The case of INSPEC records for Chilean (*) authors is most unusual in that it even increases in number of records for 1990. These are marked INC in Appendix 2.

The Chart for INSPEC records for authors from Brazil will serve as an example.

Chart 1
INSPEC RECORDS FOR DOCUMENTS WITH
AUTHORS FROM BRAZIL
Number of Records Added by Year of Publication

Eight cases can be interpreted as increasing, but peaking in 1988 and beginning a decline in 1989 (a decline beginning in 1990 is expected and common, because of delays in acquisition and processing): Aerospace Abstracts for Argentina; GeoRef records for Brazil; INSPEC records for Argentina; Life Sciences for Brazil and Chile; MATHSCI records for Argentina and Brazil; and, MEDLINE for Venezuela. These are marked ID88 in Appendix 2.

Four cases can be interpreted as increasing, but peaking in 1987 and beginning a decline in 1988: BIOSIS records for Venezuela; INSPEC records for Venezuela; NTIS records for Brazil; and, PASCAL records for Argentina. These are marked ID87 in Appendix 2.

Four cases can be interpreted as increasing, but peaking in 1986 and beginning a decline in 1987: BIOSIS records for authors from Colombia and Argentina; Life Sciences for Argentina (*); and, PASCAL for Chile (*). These are marked ID86 in Appendix 2.

Based on past personal experience in international database development, it is hard to imagine at this point that decreases are delays due to acquisitions and processing problems. Chart 2, describing the trend for Argentina in Life Sciences, appears below.

Chart 2
LIFE SCIENCES RECORDS FOR DOCUMENTS WITH
AUTHORS FROM ARGENTINA
Number of Records Added by Year of Publication

Five cases can be interpreted as increasing, but peaking in 1985 and beginning a decline in 1986: COMPENDEX records for Chile; BIOSIS records for Chile; GeoRef records for Venezuela; PASCAL records for Colombia; and, PsycINFO records for Brazil. These are marked ID85 in Appendix 2.

Three cases can be interpreted as increasing, but peaking in 1984 and beginning a decline in 1985: Aerospace Abstracts records for Chilean authors; and, PASCAL records for Brazilian and Venezuelan authors. These are marked ID84 in Appendix 2.

One case each can be interpreted as increasing, but beginning a decline between 1981 and 1983: Aerospace Abstracts records for Brazil, and Chemical Abstracts records for Chile and Brazil. These are marked ID82 in Appendix 2.

It is interesting to note that the chief increases are in BIOSIS, INSPEC, and MEDLINE: the situation is the same for Europe, although there are more cases there. In PASCAL, the bulk of the cases were of the peaking type; however, the years of peak were later in the 1980s. The same can be said for Chemical Abstracts. While this approach may appear to overwork minimal data, the classification was found to be essential in handling the European trends. It lays a foundation for comparison later among all four regions of the world.

Pure vs. applied Sciences
Overall ranks were computed for the different countries, as indicated in Appendix 2. Especially in working with developing country information, the question arises as to whether trends can be identified to differentiate pure and applied sciences. In an attempt to investigate this point, BIOSIS, Chemical Abstracts, Life Sciences and MATHSCI were assumed to favor pure sciences, and the other databases were assumed to favor applied sciences. Average ranks were computed for each category of databases. These are shown in Table 4.

Table 4
OVERALL AND AVERAGE RANKS OF COUNTRIES IN
PURE AND APPLIED SCIENCES

A Spearman’s r_s was calculated to measure the strength of association between the two ranked lists. Its value was .999, indicating that there is an extremely high correlation between the two. While there are slight visual differences, statistically they are essentially the same. No country produces more (or less) pure or applied science than any other country.

 Discussion

Interpreting Results
There are several points worth noting before discussing the findings. First, there are some searching and coding problems. While best efforts have been put forward to resolve inconsistencies in the identification of countries, in particular the case of Guyana and French Guiana, and to evaluate the data according to what is present and consistently coded as of a given year, it is virtually impossible to be absolutely certain that every appropriate record has been correctly identified. The coding problems of MATHSCI, MEDLINE, and BIOSIS are noted above.

Second, databases are constructed within organizations comprised of people who undertake actions and promulgate policies that affect the development of a file. These may positively affect the incidence of records, in the case of the establishment of a new relationship with another organization, or negatively affect incidence, for instance when records are deliberately removed in a database cleanup operation. Specific cases have been described earlier [2], [5]. These works show that databases are not in all cases the precise mirror of scientific activity that we might like them to be. Other instances also have been described [6]. Nonetheless, these are the data available, and the picture that emerges. Caution is urged in interpretation of the data, because the trends shown may not absolutely reflect the literature actually being produced.

Third, in all cases except MATHSCI, only the author affiliation of the first author is given: the affiliation of second and third authors is not given. In actuality, therefore, the real numbers of authors are a good deal higher. However, there are no data from which to extrapolate the likely contributions of co-authors. One can only assume that the nationalities of second and third authors are likely to be related to those of first authors; the data in one study weakly suggest this [7].

It cannot be said that coverage of South American authors is increasing overall in these bibliographic databases. In 51% of cases, there were insufficient data on which to hazard a guess whether coverage was increasing or decreasing, and a very liberal definition of sufficient was used. In the remaining cases for which there were data of some substance, only seven of the 34 cases showed definite increasing trends. The remaining cases were presented with caveats of declines during the later 1980s. It is worth while to note that there were no instances of definite declining patterns, although that is hardly grounds for optimism.

While it is common to think of Third World research as concentrating in the applied sciences, the picture that emerges from this examination does not support that impression. In fact, in terms of overall coverage, agriculture and medicine are the lowest in number of records included. Chemical Abstracts contains the largest number of records for documents with authors from South America. This is not really unexpected: it is also the largest database.

Brazil's coverage, at least quantitatively, is strongest, followed by Chile, Venezuela, and Argentina, and possibly Colombia.

Earlier papers noted above have interpreted the analysis of trends in bibliographic databases as indicators of scientific activity. Note that this approach bypasses the issue of South American scientists publishing in North American journals: the source of the journal itself is ignored. In this light, the picture of South American science is bleak indeed, with only a few countries maintaining active contributions to these disciplines. The picture is largely of declining activity. Earlier studies have sought to relate trends in authorship, and in other factors concerning the ‘internationality’ of these databases to specific database provider policies and practices; to the aggregate growth of the literature; and to database problems. In this case, no firm statement is clear. There is no evidence of aggressive attempts to seek out and include South American literature, other than in the cases of MEDLINE, BIOSIS, and INSPEC to a degree. The picture that emerges is more one of random chance: that if the documents find their way into the offices of the database providers, they are welcome and included. And there is no strong argument for this explanation either.

 Conclusions

South American authors, other than those from Brazil and Argentina, Venezuela, and Chile to a lesser degree, are receiving decreasing attention in the international bibliographic databases. The works of most countries are barely noticeable, in comparison with the European countries, North America, and the larger producers of Asia and Africa. Should these trends continue, the online databases will be decreasingly attractive to those seeking information produced by South American authors. It is hoped that South American files, such as the Index Medicus Latinoamericano, will become more visible as resources for United States and international awareness of South American scientific and technical activity.

 References

1. Gretchen Whitney, “Asian Authorship in Major International Bibliographic Databases: Trends for 1970-1990,” presented at the Third International Conference on Infometrics, Indian Statistical Institute and the Indian National Scientific Documentation Centre, Bangalore, August 1991.

2. Gretchen Whitney, Languages in Databases: An Analysis and Evaluation (Metuchen, N.J.: Scarecrow Press, 1990).

3. Gretchen Whitney, “Access to ThirdWorld Science in International Scientific and Technical Bibliographic Databases,” presented at the Unesco International Conference on Science Indicators for Developing Countries, Paris, October 1990.

4. Gretchen Whitney, “Science and Technology in the Gulf Region: Patterns of Authorship in Major Bibliographic Databases,” unpublished, 1991.

5. Gretchen Whitney, “Organizational Variables Affecting The Conduct of Bibliographic Database Research,” Information Processing and Management (in press).

6. Gretchen Whitney, “Patterns of Authorship in Major Bibliographic Databases: The European Region,” unpublished, 1991.

7. Gretchen Whitney, “Science and Technology in the Gulf Region: Patterns of Authorship in Major Bibliographic Databases,” unpublished, 1991.

 Appendices

Appendix 1: Countries as Ranked in Databases, by Gross Numbers of Records

Notes:
INC — Increasing trend overall for years in which data is available
TS — Too small, less than 100 records in any given year
E — Erratic or stable, no apparent trend
IDnn — ncreasing to the year specificed, but declining thereafter
e.g., ID85 — Increasing until 1985, but beginning a decline in 1986 that carries through 1989
Countries in database with no value: that country did not fall into the set of countries that made up 95% of the database.

Appendix 3: Ranges of Records in Databases for South American Authors (in 000s)

About the Author

Gretchen Whitney is an Assistant Professor in the Graduate Library School, University of Arizona.