Skip to main content
Publishing

Publishing

Blasts from the past

10 Feb 2004

It does not matter if they were published 10 years ago or 100 years ago, old scientific papers may be more important than you think, as Werner Marx and Manuel Cardona explain

Table 2

How can the significance or usefulness of a scientific paper be measured? One way to do this is to count the number of times that a given paper has been included in the reference lists of other papers. This citation approach can be applied to individual papers or scientists, and also to journals, areas of science and whole countries. However, the number of citations cannot easily be equated with the overall significance or usefulness of a paper. This is true for recent papers, the long-term significance of which may not yet be clear, and also for many older papers that are not cited because their results are now so well known that they appear in textbooks. It would be easy to theorize and speculate about these matters, but there is a much more satisfactory way to proceed: as is always the case in physics, the best way to make progress is to collect and analyse data.

In this article we look at papers from the time of Newton right up to the present day. We will concentrate on the pre-1900 and pre-1930s eras, but we will also explore some more recent trends. Most citation data come originally from the Science Citation Index (SCI) published by Thomson ISI in Philadelphia. The results in this article were obtained using SCISEARCH, the version of the SCI offered by STN International. Although SCISEARCH only covers papers published since 1974, the references in these papers obviously stretch back much further than that date.

Since 1974 only about 0.5% of the references in papers in all fields of science have been to articles published before 1900, whereas about 4% have been to papers that were published before 1950. When the “age distribution” of the references in all the papers that have been published in a particular year is plotted, it tends to peak three years previously. For example, papers published in 1999 dominate the reference lists of articles published in 2002. We can also use this plot to calculate the “half-life” of papers, a concept introduced by the crystallographer John Desmond (JD) Bernal in the late 1950s. For a given year the citing half-life is defined as the number of years that one has to go back in time to account for 50% of the total references given in that year.

We are not convinced that the increase in the citing half-life that we have found – from 9.3 years in 1975 to 10.5 years in 2002 – is significant, although a similar increase has been found in a recent, more extensive analysis by Kevin Boyack and Alex Bäcker (see further reading). The fact that papers can still have an impact several decades after they were originally published is noteworthy because it contradicts the growing belief among some information scientists and others that the lifetime of scientific publications is rapidly decreasing.

The data also show that there is considerable variation between subjects. The references in physics papers tend to be older than the average for all fields of science. For example, the share of papers with pre-1900 references is about 1%, and this increases to 6% for the pre-1930 era and 17% for the pre-1950 era. Engineers tend to cite fewer old papers (only 3% for pre-1930), whereas geoscientists tend to cite more (12%).

The impact one century after publication

Of all the pre-1800 publications that have been cited since 1974, there are about 35,000 that can be processed relatively easily using SCISEARCH. Not surprisingly, many of these early works are not journal articles but book-like publications. The most-cited scientist of this era is the Swedish botanist Carl Linnaeus (1707-1778), who has received about 4000 citations since 1974. He is followed by the Swiss mathematician Leonhard Euler, Isaac Newton and two more 18th-century entomologists – Johann Christian Fabricius and John Hunter. Other names in the top 20 include Edmond Halley, Robert Boyle, Joseph-Louis Lagrange, Robert Hooke, Rene Descartes, Charles Augustin de Coulomb and Pierre-Simon Laplace.

There are some 400,000 papers in the SCI database that contain references to papers published before 1900, and a total of about 1.2 million pre-1930 references. The citation software that is currently available is not able to analyse such large numbers, but analysis is possible if we restrict ourselves to references made in physics journals. Not surprisingly, both lists read like a “who’s who” of physicists of the day and contain many Nobel-prize winners (see tables). John William Strutt (Lord Rayleigh) is the most-cited author in the pre-1900 list, while Einstein tops the table for pre-1930.

Einstein’s three most-cited pre-1930 papers (with a total of about 2600 citations) are based on his PhD thesis and show how measurements of Brownian motion can be used to determine the size of molecules. In comparison, his famous 1905 paper on special relativity has been cited “only” 450 times since 1974, making it his fifth most-cited pre-1930 paper. However, Einstein’s most-cited article of all time is his 1935 paper with Boris Podolsky and Nathan Rosen, which has over 2000 citations. This paper – which suggests that quantum mechanics cannot offer a complete description of “physical reality” – introduced what is now known as the EPR paradox.

In second place behind Einstein in the pre-1930 list is Peter Debye, who published two influential papers on the theory of electrolytes in 1923. The other names in the top 10 are Max Born, Irving Langmuir (the only industrial physicist in either list), Lord Rayleigh, Marian Ritter von Smoluchowski, Peter Paul Ewald, James Clerk Maxwell, Hermann Weyl and Paul Dirac. Born published highly cited papers on the hydration of ions (1920) and the quantum theory of molecules (1927). Langmuir, who spent most of his career at General Electric, is best known for his 1918 paper on the adsorption of gases by solid surfaces. Von Smoluchowski’s most-cited article put forward a new theory for the coagulation of colloids in 1917.

When we look at individual papers we find that the most-cited pre-1900 article was published by the Dutch applied mathematicians Diederik Johannes Korteweg and Gustav de Vries in Philosophical Magazine in 1895. This paper, which introduced the concept of solitons, received about 600 citations in all journals (not just physics journals). In the most-cited paper for the pre-1930 era, Ewald showed how to calculate the sums of functions of the type 1/rn: such calculations are central to understanding the electric and magnetic properties of solids. This paper, which was published in Annalen der Physik in 1921, has received about 1600 citations since 1974.

Sleeping beauties and other papers

It is informative to look at the citation patterns of different highly cited papers. Some peak quickly and then continue to generate a nearly constant impact over decades after passing their maximum, while others are ignored for decades before garnering lots of citations. However, the probability that a paper that is initially uncited will prove to be such a “sleeping beauty” is very low. Indeed, quite a high proportion of papers receive few, if any, citations.

Some highly cited papers seem to have been “rediscovered” in the past two decades (see figure). For instance a Rudolf Kohlrausch paper from 1847 – which introduced the idea of stretched exponentials to explain relaxation effects – had only a minor impact until it was cited in what proved to be an influential article published by Richard Palmer of Duke University and co-workers in Physical Review Letters in 1984. Palmer et al. has now been cited by over 900 other papers, with 120 of these also citing Kohlrausch’s work. Old papers can also be rediscovered when they are cited in articles in review journals. Both the Kohlrausch paper and a 1917 paper by von Smoluchowski benefited from being cited in Reviews of Modern Physics. Ewald’s 1921 paper, on the other hand, has been widely cited since at least 1974.

Citations of influential papers in theoretical physics and chemistry often do not reach their maximum until decades after publication, and it is not unknown for some of these articles to be virtually ignored for many years or even decades. This is not unexpected: it is not easy to incorporate revolutionary ideas into established scientific concepts. Moreover, some theories and predictions cannot be fully tested when they are originally proposed due to lack of suitable equipment or data. One could say that these papers are “premature”.

The history of science also contains many examples of the scientific community being resistant to new discoveries. Hermann von Helmholtz, for instance, commiserated with Faraday about “the fact that the greatest benefactors of mankind usually do not obtain a full reward during their lifetime, and that the more time new ideas need for gaining general assent the more really original they are”. And Max Planck once famously said, “a new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents die and a new generation grows up that is familiar with it”.

Finally, we should note some of the limitations of our analysis. For instance, Gustav de Vries does not appear in table 1 because he was the second author on the paper with Korteweg. However, the majority of papers before 1930 only had a single author. Also, the actual numbers in the tables are underestimates because they only include citations from physics journals. The overall citation counts are somewhat higher because papers are often cited outside their own discipline. This is particularly true for early papers, because science was more interdisciplinary then than it is now.

Final thoughts

So why are scientists so obsessed with recent publications, often at the expense of older work? One possible explanation is that the number of papers published every year in the natural sciences has increased by a factor of between two and four since 1974, which means that there are more new papers to read – and even less time than before to reread older papers. The Web has also increased both the pace of the publishing process and the volume of material published. It is obviously important to stay up to date with the latest research, but not at the expense of all the papers that have gone before.

Related events

Copyright © 2024 by IOP Publishing Ltd and individual contributors