Scientometrics, Vol. 2, No. 1 (198) 53-63 THE SUBJECT COMPOSITION OF THE WORLD'S SCIENTIFIC JOURNALS M. P. CARPENTER, F. NARIN Computer Horizons, Inc., 15 Kings Highway North, Cherry Hill, New Jersey 834 (USA) (Received March 7, 1979) A country by subject count of the serial periodical collection at the British Library Lending Division (BLLD) in 1973 is reported and compared to previous counts. Approximately 25 periodicals have titles indicating that they are scientific journals in nine fields of the physical and biological sciences, engineering, and mathematics. The overall subject distribution of the journals appears to be remarkably stable when compared to a similar count by Hulme 6 years ago, although the number of journals appears to have doubled in the last 6 years. A major shift was found in the national origin of the journals, when compared with Hulme's counts, with a notable rise in the number and percent of U.S. journals, and a sharp decline in the percentage of French and German journals. Introduction The size of the world's scientific literature has long been of interest to scientists, to historians of science, a'nd to members of the information science community. As part of our work for the National Science Foundation we developed national publication and citation comparisons, based on country and subject classification of the papers in journals covered by the Science Citation Index (SCI). 1-3 This paper describes analogous comparisons based on counts of journals acquired by the British Library Lending Division (BLLD). There are two commonly used methods of measuring literature size: article counts and journal (or serial periodical) counts. An article count is the more desirable but also the more difficult to make. A journal count is substantially easier to make but comparisons of journal counts for different countries are subject to two sources of bias: all journals are not of equal size, and all papers in a journal are not from the country of origin of the journal. As a result of these limitations, national activity as measured by journal counts will tend to be biased against the major scientific countries, from which most of the larger journals originate. This bias is somewhat ameliorated by the fact that many scientists from less scientifically developed countries are publishing in major Scientometrics 2 (198) 53
M. P. CARPENTER, F. NAP, IN: SUBJECT COMPOSITION OF JOURNALS country journals. For example, our recent work for the 1976 Science Indicators report indicates that 25% of the papers in U.S. mainstream joumals are authored by scientists in other countries, while only 18% of U.S. authored papers are published in foreign journals. 4 For India, a less scientifically developed country, only 5% of the papers in SCI covered journals published in India are authored by scientists from outside of India, whereas 63% of Indian authored papers in the SC1 appear in journals published by other countries. In an earlier paper on national publication and citation comparisons we reviewed earlier article counts and showed that the relative publication rates of the major countries of the world had changed radically "when viewed in a perspective covering the 2th Century, during which there had been an enormous rise in U.S. and U.S.S.R. publications, and a striking relative decrease in German and French scientific publications". 2 We will show in this paper that the same phenomenon is evident in counts of scientific journals. BLLD serials Construction of the data base In order to deal with the most complete collection of currently published scientific serials, we contacted the British Library Lending Division (BLLD) and obtained from them a magnetic tape of all serial publications received at the BLLD in 1973. s The BLLD attempts exhaustive coverage of the world's scientific litera- ture. One particularly attractive feature of using the BLLD for this study is the fact the list is constantly culled, and serials which ceased to be received by the BLLD are removed. An excellent discussion of the scientific and technological periodicals collection at BLLD is contained in a paper by Ban', describing the collection of serials at the National Lending Library for Science and Technology (NLL, now part of BLLD) in 1965. 6 The magnetic tape acquired from BLLD included the "country" of origin of each serial. There were 81 different countries, territories and geographical regions on the 1973 tape. The data presented in this paper are grouped into a set of 14 countries and six geopolitical regions, corresponding to the groupings used in our Science Indicators work. This grouping includes the 1 countries whose scientists author the largest number of articles in the Science Citation Index data base: U.S., U.K., U.S.S.R., Germany, France, Japan, Canada, India, Italy and Sweden. Four 54 Scientometrics 2 (198)
other countries are included separately because they would dominate their regions if left therein: Israel, South Africa, Australia, and New Zealand. The remaining countries of the world are assigned to appropriate geopolitical regions. Assignment of serials to fields Based on its title, each of the serials on the tape was assigned to one of the following nine basic fields of science, if its title so indicated: 1. clinical medicine, 2. biomedical research, 3. biology, 4. chemistry, 5. physics, 6. earth and space science, 7. psychology, 8. mathematics, 9. engineering and technology. It is, at times, difficult to assign a serial to a field based on title alone. Only 59% of the 43 457 serials were considered to fall within these nine basic fields, since many of the serials appeared to be in the social science or humanities. Other serials not assigned to the nine fields were considered to be hobby journals, statistical compendia, news periodicals, trade journals and other periodicals which would not ordinarily be considered scientific research journals. Our final count of 24 81 journals for the nine fields in the BLLD collection for 1973 is remarkably close to Barr's count of 25 75 serials in the NLL collection in 1965; NLL defined science and technology as the natural, physical and engineering sciences, excluding social science, but including psychology and some borderline subjects. The agreement between Barr's count and our count indicates that the assignment based only on title was successful in extracting from the full BLLD set the fields of interest. Many of the titles were not English language, and although many scientific words are similar from language to language it was still necessary to obtain the assistance of translators for titles in the following languages: Czech, Danish, Dutch, Finnish, French, German, Hungarian, Italian, Japanese, Polish, Rumanian, Russian, Spanish, Swedish, Turkish and the languages of Yugoslavia. In the end, 9.3% of the titles were not specific enough to assign to a field. Most of these are probably not scientific journals. A country by country check of the number of joumals which could not be field assigned did not reveal any apparent serious country bias in unassigned titles. Further, our overall country distribution is quite close to the distribution found by Bart. Scientometrics 2 H98) 55
Results The results of the count are contained in Table 1, a country by field cognt of journals which were assigned to the nine basic fields for 1973. The United States publishes more journals than any other country, in every one of the nine fields. Overall, the United States is the country of origin of 2% of the BLLD journals. For comparison, our paper counts based on the 1973 Science Citation Index data indicate that the U.S. is the source country for 37% of the world's scientific papers. The United Kingdom is second to the U.S. in this BLLD count; in contrast, counts of scientific articles based on the Science Citation Index indicate that the Soviets produce more scientific papers than the U.K. However, the location of BLLD in the U.K., and the fact that it obtains periodicals mainly on the requests of British scientists, could provide an understandable bias toward British periodicals. The data in Table 1 may also be viewed in terms of the percent of each field's journals which are published in each country. These percentages are given in Table 2. The first observation is that the U.S. published between 15 and 25% of the journals in every field, with the exception of psychology, for which 4% of the journals are of U.S. origin. An additional 3% of the psychology journals covered by BLLD are from the U.K., Germany, or France, while only 1% are from the Soviet Union. For most countries and geopolitical groups there is little variation in the percents of the fields produced by the country. The French seem to publish a comparatively small fraction of the chemistry journals, while the Soviets publish a rather large fraction of physics journals, compared to other fields. The relatively high percentage for Australia and New Zealand in biology reflects a propensity in those countries towards publication in the food and agriculture areas which are important to their economies. The overall patterns of national publication of papers, based on SCI data, are discussed much more extensively in our recent paper on the distribution of world science. 7 Comparison with previous counts An interesting use of the data presented in Tables 1 and 2 is to compare it to earlier studies of the scientific literature. A particularly fruitful comparison involves the set of lectures on statistical bibliography which Hulme delivered at the University of Cambridge in 1921. 8 By aggregating various fields and countries, comparisons can be made between the Hulme data and the BLLD data. 56 Scientometrics 2 [198]
.~ ~'~ ~ ~ ~-,.-4,...4 ~ ~ o~,~,.4 r m.,-4 r ~a ~ ~, Z / / / Scientometric: 2 (198) 57
M, P, CARPENTER, F. NARIN: SUBJECT COMPOSITION OF JOURNALS 9 ~,4 i~ ~~o.~ ~ o ~ ~~ o o o~ 58 Scientometrics 2 (198)
Table 3 Field-by-field comparison of Hulme's counts with our BLLD* counts Hulme BLLD* Field (198--1912), % (1973), % Biomedicine Biology Chemistx3r Physics Earth&Space Science Mathematics Number of jouxnals 22.8 36.5 6.6 7.7 23.7 2.6 7 61 35. 29.7 5.9 5.9 18.2 5.2 16 346 *Excluding Engineering&Technology and Psychology to correspond to Hulme. For a comparison with Hulme's data in terms of the world's journals by field, we have taken his table of all journals referred to in annual issues of the International Catalog of Scientific Literature, averaged for the years 198 through 1912. We then aggregated Hulme's subject categories into the following six fields: biomedicine, biology, chemistry, physics, mathematics, and earth & space science. Similarly, the BLLD data were aggregated into the same six fields, by putting clinical medicine and biomedical research together, and by omitting psychology and engineering & technology. Table 3 compares the two counts, when they are so aggregated. One exceedingly interesting observation is that the BLLD count for 1973, some 63 years later than the Hulme count, reveals that the number of journals in these fields have little more than doubled, rising to 16 346 from 7 61, a growth of 114%. If we assume an exponential rate of growth of journals between 191 and 1973, then the doubling time for surviving scientific journals appears to be about 57 years, which is much longer than the doubling time for the founding of scientific periodicals, approximately 15 years, as noted by Derek Price. 9 Another direct observation from the table is that the relative proportions in the different fields have not changed very greatly over the years. Biomedicine and biology (for which the distinction is very fuzzy) are still the fields with the largest numbers of journals. The percentage of earth and space science journals varies only slightly and remain rather large - almost certainly due to the large numbers of relatively small journals in such fields as meteorology, mineralogy, geology, and geography. Scientometrics 2 (198) 59
Table 4 Countr7 comparison ofhulme's counts with our BLLD* counts Hulme** * % [ BLLD* Country (191-1913), (1973), % US UK Germany France USSR Japan Australia Canada India Italy Sweden New Zealand South Africa Rest of Europe Rest of Asia Rest of Africa Middle, Near East& North Africa Central&South America Number of journals 7.9 7.8 28.6 12.2 9.6 1.6 1..9.6 7.4 1.3.1.6 19.7.1.2.6 8 228 19.6 1.6 7.5 5.1 8.5 6.1 2. 2.6 1.9 3.9 1.5.6 1.9 19.8 2.1 1.4.8 5.1 16 346 *Excluding Engineering&Technology and Psychology to correspond to Hulme. **Data on all journals from 191-1913. In contrast to the field-by-field stability, a comparison of Hulme's counts on a country-by-country basis with our BLLD based counts reveals great changes in this century, with much the same pattern we noted in our earlier paper) Table 4 shows this comparison. The most striking changes occurring between the first decade of this century and the 197s in the country of origin of scientific serials are the rise in the U.S. fractions and the decrease in the German and French fractions. Also notable is the substantial increase in the percentage of Japanese joumals. The smaller countries of Africa, Asia, and Central and South America now publish al- most 1% of the world's scientific journals, compared to barely 1% 6 years ago. 6 Scientometrics 21(198)
M. P. CARPENTER. F, NARIN: SUBJECT COMPOSITION OF JOURNALS Table 5 Country comparison of Gottschalk and Desmond's counts with our BLLD counts Country US UK Germany France USSR Japan Australia Canada India Italy Sweden New Zealand Rest of Europe Rest of Asia South Africa and Rest of Africa Central&South America Others (not identified) Israel Israel Rest of Near East and North Africa Number of journals Gottschalk and Desmond (1962-estimated), % 17.6 6.2 8.7 7.9 6.2 7.9 1.3 1.6 1.8 4.3 2..4 19.7 3.8 1.8 7.5 1.3 35 3 BLLD counts, % 2.1 12.6 8. 5.2 8.8 6.1 2.1 2.4 1.9 3.4 1.7.6 19.1 1.8 1.8 4..2.4 24 81 A recent rise in scientific publications from the smaller countries is also apparent in counts of scientific papers. It should be noted that our article counts of Soviet science indicate a much larger U.S.S.R. role in the 197s than indicated by these BLLD journal counts. Another comparison with an earlier count is shown in Table 5, the world-wide census of scientific and technical serials estimated by Gottschalk and Desmond ha ]962.1~ Their total count is substantially larger than the BLLD counts for a number of reasons. They strove for comprehensiveness and included borderline areas, such as anthropometry, as well as the central natural, physical and engineer- Scientometrics 2 (198) 61
Lug sciences. Further, and more significantly, they obtained their counts from national listings of current titles. As a result, man3~ of the titles listed were probably not in current publication. The net result is that their count is about 4% higher than our BLLD count and Barr's NLL count in 1963. They found a much smaller percentage for the U.K. and a somewhat smaller percentage for the U.S. They found larger percentages for France, and a smaller percentage for U.S.S.R. Overall, though, it is clear that their country distribution resembles the BLLD country distribution much more closely than it resembles Hulme's country distribution of 6 years ago. Various other counts with which our's might be compared are contained in the 1962 article by Bourne which estimated the origin, language, field, indexing and abstracting of the world's technical journal literature, based on queries to many different individual sources. 11 Bourne's data are mainly a collection of the estim- ates of others, rather than original data. Conclusion The counts presented here indicate that the number of scientific serials has doubled in the last 6 years, with a substantial change in national distribution but a relatively small change in subject distribution. The rise found for U.S. serials and decline in German and French serials parallels earlier findings in terms of scientific papers. Work supported under the National Science Foundation's Contracts #SRS 77-2277 and C-896. The authors wish to acknowledge the advice and assistance of Dr. Robert Wright of NSF in the performance of this work. Notes and references 1. National Science Board, Science Indicators 1972, Washington, D. C., 1973. National Science Board, Science Indicators 19 74, Washington, D. C., 1975. 2. F. NARIN, M. CARPENTER, National Publication and Citation Comparisons. Journal of the American Society for Information Science, 26 (1975) No. 2, 8-93. 3. Science Citation lndex, SC1. Copyright Institute for Scientific Information, Philadelphia, PA. 4. National Science Board, Science Indicators 1976, Washington, D. C., 1977.. 5. The tape was obtained from BLLD by D. B. McCARN, National Library of Medicine, and forwarded to us by permission of Dr. A. HARLEY of the British Lending Library. 6. K. P. BARR, Estimates of the Number of Currently Available Scientific and Technical Periodicals. Journal of Documentation, 23 (1967)No. 2, 11-116. 62 Scientometrics 2 (198)
7. J. D. FRAME, F. NARIN, M. P. CARPENTER, The Distribution of World Science. Social Studies of Science, 7 (1977) No. 4, 51-516. 8. E. W. HULME, Statistical Bibliography in Relation to the Growth of Modern Civilization, London, Grafton, 1923. 9. D. de SOLLA PRICE, Little Science, Big Science, New Haven, Yale University Press, 1963. 1. C. M. GOTTSCHALK, W. F. DESMOND, World-Wide Census of Scientific and Technical Serials. American Documentation, (July, 1963) 188-192. 11. C. P. BOURNE, The World's Journal Literature: An Estimate of Volume, Origin, Language, Field, Indexing, and Abstracting. American Documentation, (April, 1962) 159-168. Scientometrics 2 (198) 63