When I was a PhD student in the 1970s, and in fact for a long time after as a researcher and academic, to keep up to date with new literature it was necessary to go to a library holding the journals of interest. In my case the primary journals of interest included Water Resources Research, Journal of Hydrology, Hydrological Sciences Bulletin (as it was then), Water Resources Bulletin, ASCE Journal of Hydrologic Engineering, Journal of Soil Science, Soil Science Society of America Proceedings, Journal of Agricultural Meteorology and Quarterly Journal of the Royal Meteorological Society (Hydrological Processes, of course, did not exist until much later in 1987). That was also the way of browsing old content, and in about 1980, while browsing in old issues of the Transactions of the American Geophysical Union, it was how I came to learn about the 94 boxes of Robert Horton's papers that have been deposited in the US National Archives (see Beven 2004). Robert Elmer Horton (1875–1945) was born 150 years ago in May this year, and when I was a doctoral student, he seemed to be only part of hydrological history, but it is curious to note that he died only 25 years before I started my research career (i.e., the same period of time since the change of millennium!).

For new journal content in the library, there were often special shelves for new issues, usually with the copies for the current year stored behind. Sometimes it was necessary to go to the shelves to check on issues before they went for binding into annual (or more frequent in the case of Journal of Hydrology) volumes. There were already some abstract systems (in hydrology and geomorphology Keith Clayton at the University of East Anglia had founded Geoabstracts in 1969) but these were still hard copy and necessarily came out sometime after the journals themselves. So it was more effective, if somewhat time-consuming, to spend time in the library checking on new papers in the journals of interest as they became available.

This had two important advantages. One was that of serendipity. In going through the contents pages of journals in hard copy, it was often the case that you would find interesting papers outside of your main interests. Sometimes this could lead to following up an interesting topic that might later lead to some useful outcome (as was the case with the Horton archive). The equivalent today is browsing through the emails of journal alerts, but the number of emails is ever increasing, the lists of paper titles and abstracts are ever longer and there is naturally a tendency to filter out anything that does not appear immediately relevant rather than clicking through to an abstract or paper. The second advantage was that, reading a paper in hard copy, you could be sure that it existed. That is not always the case for papers found by internet search engines such as Google Scholar or reports from artificial intelligence (AI) agents.

This brings us to the curious case of the paper of Robert Horton in the Transactions of the American Geophysical Union (TAGU) in 1940. The Transactions at that time served to publish the reports of the various Committees and Subcommittees of the different Sections of AGU, as well as the written version of papers presented at meetings of AGU each year. The AGU Section of Hydrology, which had been founded only 7 years earlier in 1931, had multiple Committees in 1940. Reports were presented for the Committees on Runoff (by W.G. Hoyt), Snow (by J.E. Church), Glaciers (by François E. Matthes), Evaporation and Transpiration (by Joseph Kittredge), Infiltration (by G.W. Musgrave), Physics of Soil Moisture (by N.E. Edlefsen), Underground Waters (by David G. Thompson), Physical Limnology (by T.C. Adams) and Dynamics of Streams (by Lorenz G. Straub). There were also reports from the Special Committee on Flood Waves (also by Lorenz G. Straub), the Coordinating Committee between the American Society of Agricultural Engineers and AGU (by H.S. Riesbol) and the Advisory Committee of the Soil Conservation Service, written by Robert Horton for which he served as Chair (reports were presented during the years 1938–1940).

Some things do not change much with time, it seems. Robert Horton also has two papers listed that year, one with Edward L. Beutner and Ralph R. Gaebe on ‘Sprinkled-Plat Runoff- and Infiltration-Experiments on Arizona Desert Soils’ (pages 550–558), and one as sole author with the title ‘The infiltration-theory of surface runoff’ (p541). The reason why only one page is listed for the latter is shown in Figure 1.

There are, of course, many other papers by Horton relating to infiltration (see, for example, Beven 2004, 2021). There is even a paper by LeRoy K. Sherman some years earlier with the same title, as an Appendix to the Report from the Committee on Rainfall and Runoff 1934–1935 (Sherman 1935) which aimed to summarise Horton's concepts.

The reason why Horton's (1940) paper might be considered curious is, however, that Google Scholar suggests that it has been cited 25 times (i.e., more than over 100 of my own papers on Web of Science that I know to contain content, even if they are clearly of limited interest to others). Several of these citations are marked by Google Scholar as [citation] including four with dates before 1940 (not sure what the Google Scholar algorithm is doing there), and one in 1942 in the volume by H.W. Wood on Flood flow on Missouri Streams, published by the University of Missouri (Wood 1942). This is available online and does indeed cite Horton (1940).

However, much more curiously, the remaining 21 citations are all after 2016, when it is cited in a paper in the Italian Journal of Groundwater. In 2017 there is a citation in the Journal of Sustainable Water in the Built Environment, but then things start to get more mainstream with citations in Vadose Zone Journal, Environmental Pollution, Journal of Hydrologic Engineering and Journal of Irrigation and Drainage Engineering. It even appears in a Handbook of Water Resources Engineering and a Handbook of Atmospheric Measurements. There are already citations listed in 2024 and 2025. Clearly, the title suggests that this paper gives a good summary of Horton's infiltration theory, but the authors have not actually checked (and, of course, such a citation would also seem to be entirely plausible to referees).

So the moral of this story is that you should not trust Google Scholar or AI agents to provide useful material based on titles alone. It is certainly an efficient way of filtering the ever-increasing volume of hydrological literature when the pressure is to produce outcomes ever faster. Just be aware that sometimes this might be a matter of working too fast if it means using search results without actually checking if a reference produces what the title and citation suggest. This is not totally new, of course. There was always the possibility in the pre-internet era of taking titles from reference lists in published papers and citing those papers without checking. But the danger is clearly greater today when journal contents lists and Google Scholar searches can be delivered directly to the screens on our desks or even to our mobiles. There are already documented cases of AI agents ‘hallucinating’ spurious references to support answers to queries (and even inventing fictional legal cases to support a position—which might be considered even more serious). Perhaps, as both authors and referees, despite the pressures to work ever faster, we should slow down just a little and take a little more care in our citations.