Changing U.S. Output of Scientific Articles: 1988-2003
Output Trends in Major S&E Publishing Centers
Trends in U.S. Article Output
The number of S&E articles credited to U.S. institutions through fractional counting of the expanding Thomson ISI journal set has essentially remained constant since 1992 after growing consistently during the preceding two decades. It is convenient to divide the 1988–2003 period into an early period (1988–92) when U.S. article output was growing and a later period (1992–2003) when it was essentially flat. Output grew at an average annual rate of 0.6% between 1992 and 2003 but all the growth occurred in 2003, so growth was -0.2% through 2002, compared with 2.9% between 1988 and 1992 (figure 1 ). Appendix table 3 shows U.S. article output and world article share based on fractional counts.
The flattening of U.S. article output occurred despite continued growth of both real R&D expenditures and the academic sector R&D workforce, which produces most U.S. S&E articles (figure 2 ). Article output of the three other major S&E publishing centers (the EU-15, Japan, and the East Asia-4) grew considerably faster than U.S. output between 1992 and 2003 (figure 1 ). The growth rates of the EU-15 and Japan during this period, however, were considerably slower than between 1988 and 1992, in a pattern similar to that of the United States.
These changes are not the result of changes in Thomson ISI journal coverage. Although expanding journal coverage over time could affect measures of U.S. output for reasons that have little or nothing to do with publishing intensity, such as coverage of new non-English journals, data from the fixed set of journals that have been part of the Thomson ISI database since 1985 show a similar growth pattern and identical change in the U.S. share (appendix table 2). In several other journal databases, the trend of faster growth in the rest of the world relative to the United States was also similar (appendix table 1). Therefore, except as otherwise noted, this section reports data for the Thomson ISI expanding journal set.
Throughout the 1988–2003 period, international collaboration as indicated by article coauthorship was rising, which affects whole and fractional authorship counts differently (see sidebar "Effect of International Collaboration on Whole- and Fractional-Count Output"). However, overall U.S. output trends based on these two measures are congruent, albeit with somewhat higher U.S. growth rates when whole counts are used. Thus, U.S. whole-count output slowed in a pattern similar to the flattening of output seen when fractional counts are used (figure 3 and appendix table 4), rising at an average annual rate of 1.2% between 1992 and 2003 compared with 3.4% between 1988 and 1992 (figure 3 ). Output growth in whole counts in the three other major publishing centers between 1995 and 2003 was higher than in the United States.
Unless otherwise indicated, the remainder of this report presents fractional-count data for the expanding Thomson ISI journal set. However growth trends by field seen in the whole-count data and in the fixed journal set are similar.
U.S. growth trends by field showed some variation (appendix tables 3 and 5). In some fields, the pattern tracks the overall trend fairly closely, but with different dates for the onset of flattening. These include the social sciences (1990), biology (1990), engineering/technology (1993), chemistry (1993), clinical medicine (1995), biomedical research (1995), and psychology (1996). In physics, article output flattened at the same time as overall output but exhibited an outright decline between 1995 and 2003.
In two fields, trends departed substantially from the overall pattern. Article output in the earth/space sciences grew consistently over the entire 15-year period. In mathematics, output declined between 1988 and 1997 but then grew modestly between 1997 and 2003.
Trends in federal funding by field did not generally coincide with trends in article output in broadly comparable fields (table 1 ). Beginning in 1990, real growth in federal funds was generally positive and robust, except in chemistry and physics. Yet corresponding growth in article output occurred only in the earth/space sciences, which receive funds classified under geosciences and astronomy. In physics, the funding trend—an increase in real terms between FY 1986 and FY 1990, followed by a decline—was also similar to the trend in article output.
U.S. Share of the World's Articles
The U.S. share of world article output declined between 1992 and 2003. The combination of nearly stagnant U.S. output and continued growth in the three other major S&E publishing centers led the U.S. share to fall from 37% to 30% (figure 4 and appendix table 3). A longer series that is limited to natural sciences and engineering articles shows a long term decline from 38% in 1973 to 28% in 2003 (see inset to figure 4 ).
As with output trends, share trends by field varied from the overall average (figure 5 ). Two fields, engineering/technology and mathematics, experienced percentage point declines that were more than twice the overall U.S. decline during this period. Biomedical research, clinical medicine, the earth/space sciences, and chemistry had a significantly lower than average decline in their shares compared with the overall share trend.
For most fields, the size of the share decline was similar whether measured in whole or fractional counts (figure 5 and appendix tables 3 and 4). Three fields, clinical medicine, biomedical research, and the earth/space sciences, were exceptions. In these fields, the decline in whole-count share was substantially less than the decline in fractional-count share. This finding suggests that, more than in other fields, the loss of fractional-count share in these fields was because U.S. authors were increasing their collaboration with the rest of the world (see sidebar "Effect of International Collaboration on Whole- and Fractional-Count Output").
Article Output in the EU-15, Japan, and the East Asia-4
In contrast to the flattening of article output in the United States between 1992 and 2003, output expanded during this period in the three other major S&E publishing centers (figure 6 and appendix table 3).
Article output in the EU-15 grew at an average annual rate of 2.8% between 1992 and 2003, more than four times faster than in the United States (figure 1 ), and EU-15 output surpassed U.S. output in 1998 (figure 6 and appendix table 3). As in the United States, EU-15 growth after 1992 was slower than growth between 1988 and 1992. The EU-15 growth trend in the 1985 fixed journal set was very similar and surpassed U.S. output in 1997 (appendix table 2). EU-15 fractional-count output grew faster than U.S. output between 1992 and 2003 in all fields (appendix table 5). In the mid-1990s, EU-15 output surpassed the United States in clinical medicine, the earth/space sciences, engineering/technology, mathematics, and physics (appendix table 3).
Japan's output rose at an average annual rate of 3.1% between 1992 and 2003, five times faster than that of the United States (figure 1 ). As with the United States and the EU-15, Japan's growth rate began to slow in 1992. Japan's output path was similar in the 1985 fixed journal set (appendix table 2). Japanese output growth exceeded that of the United States in all fields between 1992 and 2003 (appendix table 5). In physics, the gap in article output between Japan and the United States narrowed substantially between 1992 and 2003 (appendix table 3), with Japan producing less than one-half the number of physics articles as the United States in 1992, but two-thirds as many in 2003.
Authors based in the East Asia-4 produced S&E articles at a sharply accelerating pace. Between 1992 and 2003, growth in collective article output averaged almost 16% per annum, more than 25 times faster than in the United States (figure 1 ). Rapid growth pushed this region's share of the world's S&E articles from 2% in 1992 to 8% in 2003, nearly equivalent to Japan's share (figure 4 and appendix table 3). Growth in East Asia-4 article output in the 1985 fixed journal set was equally rapid, resulting in a share increase of similar magnitude (appendix table 2). Engineering/technology and chemistry were among the fields that grew the most rapidly during this period. In 1992, the East Asia-4's article output in engineering/technology and in chemistry was less than 20% that of the United States. In 2003, they reached more than 70% of the U.S. total (appendix table 3). Output in both fields surpassed that of Japan in about 2000.
 U.S. article output rose almost 8% in 2003 compared with 2002, following a decline between 2001 and 2002. A similar trend occurred in the EU countries and Japan. Although this trend may reflect a real change in article output, it may also be the result of variation in how quickly publications are added to the Thomson ISI database.
 Data on S&E article output from earlier in the 1980s and the 1970s are available in appendix table 5-44 of Science and Engineering Indicators 1998 and appendix table 5-21 of Science and Engineering Indicators 1993, respectively.
 Because the time needed to carry out research causes a lag between funding and publication of findings, the comparison of funding with article output allows for a 2-year lag between the date of funding and the date of publication (i.e., funding in year 1 is compared with article output in year 3).
 Geosciences consists of earth, atmospheric, and ocean sciences.