The Lifetimes of Astronomical Papers and the Completeness of the ADS
By counting the numbers of citations received after publication, we can determine how long astronomical papers are remembered. Those from 1955 have a half-life of 71 years and in 1960 it was 25 years. After 1970 the half-life has been a steady 10 years, telling us that astronomical results are now appreciated quickly and then replaced quickly. However, average astronomical papers have been receiving increased total citations. Tests for astronomical journals showed that it has a completeness of 94%, although it does not include most observatory publications and books
Keywords: astronomical databases — publications, bibliography — sociology of astronomy
We wonder how long our papers will be remembered. We realize that measurements that are not time-sensitive can be replaced with more accurate ones later. However, time-sensitive observations of variable stars, for instance, may be needed for centuries. We also know that certain methods need be done only once, such as the conversion from projected to spherical rotational velocities , so that paper will be cited forever and later repetitions are not needed. We also know that certain fundamental studies are useful for many decades. We also know that there are “sleepers” whose importance will not be realized for decades, such as Zwicky’s discovery of dark matter . No other person cited his paper for 27 years, but there have now been 1063 citations to it. However, sleepers are rare. We also know that contemporary communication and data recovery has become so fast that improved studies may occur more quickly. Therefore, what is the lifetime of average papers?
I looked at the citation numbers of the first 100 papers published in the Astrophysical Journal (ApJ) in the years 1955, 1960, 1970, 1980, 1990, and 2000. Papers more recent than 2000 have not had enough years to determine lifetimes. We learned that the changes in lifetimes have been drastic.
I looked at the first 100 papers published in the ApJ in each selected year. Not included were obituaries, Notes, and Letters. The Astrophysical Data System (ADS) was used to count citations. The ADS is incorrect in one way that does not affect citation counts. The ADS scans tables of contents and includes as papers the headings such as “Notes”, “Errata”, and “Reviews” as “no-author papers” whereas they are not papers at all. For instance, it states that the 1950 ApJ had 141 papers whereas the correct number is 133 papers. That affects paper totals but not citations to them.
I counted citations starting in 1955 and at decade intervals starting in 1960. The half-lives were long in 1955 and 1960 but short thereafter. One paper titled “The Luminosity Function and Stellar Evolution” has had so many citations to date (6113) that it would dominate the statistics for that year and has not been included.
These statistics are limited to 1955 and later because in that year the AAS journals (Astronomical Journal or AJ and ApJ) adopted the Harvard reference system, in which references in the texts are by author’s last names and years and the full information is given in reference lists at the ends of papers. Before 1955 authors used superscript numbers in the texts and partial information at the bottoms of pages, but not the initials of authors so the references were ambiguous. Therefore the ADS is substantially incomplete for papers before 1955. The MNRAS used superscript numbers in the texts but gave the author’s initials in the footnotes, so it avoided that problem.
The long half-lives in the 1955 are true both for fundamental papers, such as Oort & Spitzer’s “Acceleration of Interstellar Clouds by O-type Stars” , Cameron’s “Origin of Anomalous Abundances of the Elements in Giant Stars”  and Babcock & Babcock’s “The Sun’s Magnetic Field, 1952-1954” , as well as papers on individual stars producing few citations. In other words, it applies to fundamental papers as well as minor ones.
Figures 1–5 give samples of these curves. In each case I fitted the data with the best curve, usually polynomial functions. The fitted curves have no obvious significance, except to compute the half-lives.
We see that the first curve had a distribution that was essentially linear for 70 or more years. The half-lives, defined as the intervals from the maximum citations to half that number, are listed in Table 1 and are shown in Figure 6. We see that initially the half-lives were 71 yr, but by 1970 they were a steady 10 yr.
Are these results peculiar to the ApJ? I made similar counts for the Monthly Notices of the Royal Astronomical Society (MNRAS) in 1955 and 1995. The half-lives of 18.6 and 9.9 yr approximately fit the curve. I did the same for the Physical Review in 1955, 1965, 1975, and 1985. Their half-lives show a decrease with time but not as marked.
Finally, the last column of Table 1 gives the average numbers of citations per paper until 2019. It shows an increase with time, even though the times become shorter. That means that our papers receive increasing numbers of citations on average from 47 in 1955-1970 to 62 in 1980 and later. The mean total citations for MNRAS changed from 29.2 citations in 1955 to 54.5 citations in 1995. In the case of the Physical Review the total citations up to 2019 decreased from 80.7 for 1955 papers to 29.1 for 1985 papers.
We conclude that the process of astronomical research has become much faster than in the past. Astronomers quickly realize the importance of new results and make use of them much more quickly than before.
The Completeness of the ADS
These results were obtained from the Astrophysical Data System, which initially purchased data from the Science Citation Index. The ADS has the warning “The Citation database in the ADS is NOT complete”. But how incomplete is this database? We know that it does not include most observatory publications and most monographs (books), but is it complete for journal papers? We know  that journal papers now account for 90% of all cited astronomical publications. If the ADS has a completeness of 90-95%, results like the ones in Section 2 above are still valid. If it has an completeness of ~50%, its value depends on whether results from its use differs by much more than that amount.
I ran two tests for the completeness of these data. In the first test I selected five papers published in 1950 and searched the 1970 journals for citations to them. The five papers are listed in the first column of Table 2. The lead authors all had family names starting with “W” so that I needed to look only at the W sections of reference lists for reference list that were arranged in alphabetical order. For other reference lists I had to scan the entire lists. However, those five papers included three of the four most-cited papers in 1950. Then I scanned the reference list of all of the 1970 papers in the journals listed in Table 3.
The results shown in the last two columns of Table 2 tell us that of the 20 citing papers in 1970, the ADS listed 19, for a completeness of 95%. The one paper missed by ADS was by Thompson & Colvin , which cited Wilson . Perhaps that paper was missed because the authors were listed at the end of the paper, rather than below the title. The numbers in Table 2 are statistically small but involved a large amount of searching.
The second test involved counting the citations in 1982 to five papers published in 1980. Those five are listed in Table 4. I scanned all the 1982 reference lists in Table 3 but added the Publications of the Astronomical Society of Japan and did not scan Solar Physics or Solar System Research.
We see that the ADS did not cite one paper by Readhead et al. , namely Barbieri et al. , probably because that paper had two reference lists and the ADS ignored the second one. Regarding the Richstone et al.  paper, the ADS failed to cite Puetter et al. , probably because the citing paper used the wrong page number; it failed to cite Barbieri et al.  for the reason give above. The ADS failed to cite one reference to Rodriguez , namely Ambartsumian et al. , probably because that paper was written in the Russian language. In summary, the ADS failed to cite four papers out of 62 for a completeness of 93.6%. Combining the two tests, the ADS completeness was 77 out of 82 papers or 93.9%. This fraction is sufficiently large that the results above in Section 2 are valid and the ADS can be used with an accuracy of ±6% for journal papers after 1955.
This research made use of NASA’s Astrophysical Data System.
Table 1. Summary of Half-lives
Table 2. Results from Scanning Citations to 1950 Papers in 1970
The 1950 papers
Whipple (1950) 
O. Wilson (1950) 
R. Wilson & Joy (1950) 
Wrubel (1950) 
Wyatt & Whipple (1950) 
Table 3. Journals Scanned
Astronomy & Astrophysics
Astronomy & Astrophysics Supplements
Annual Reviews of Astronomy & Astrophysics
Astrophysics Journal Supplements
Astrophysics & Space Science
Journal of Geophysical Research
Journal of the Royal Astronomical Society of Canada
Memoirs of the Royal Astronomical Society
Memorie Della Societa Astronomica Italia
Monthly Notices of the Royal Astronomical Society
Planetary & Space Science
Publications of the Astronomical Society of the Pacific
Solar System Research
Space Science Reviews
Table 4. Results from Scanning Citations to 1980 Papers in 1982
The 1980 papers
Readhead et al. (1980) 
Ridgway et al. (1980) 
Richstone et al. (1980) 
Rappaport et al. (1980) 
Rodriguez et al. (1980)