Over the past few weeks, I have been working on a project to topic model the American Journal of Science between 1819 (its first year of publication) and 1922; this journal, during much of the 19th century, was the only specialized scientific journal in the United STates. I can release data sets later, but just wanted to share some preliminary results. Though this research is far from conclusive, it does provide a useful proof of concept for the method of using topic modelling to determine how genres of material change over a long period of time. Moreover, understanding this evolution of topics within a single important journal in 19th century America, helps to understand how these topics can provide a useful source of evidence to supplement more traditional historical and “close reading” methods.
The above graph shows that over the entire roughly 100 year period, topics discussing geology are the most dominant topic over time, representing roughly 35% of topics between 1819 and 1922. Interestingly, however, the “other sciences” are also represented equally at 35%. Yet, no one of the subtopics within “other sciences” dominates. Astronomy, Botany, Engineering, Medicine, Meteorology, Physics, and Zoology, individually represent less than 10% of whole. In any given year, none of these topics represent more than 13%, physics being the only exception which represents 17.5% of topics in 1840. Chemistry is one major exception. As a discipline, it represents 13% of the total topics over this 100 year period, and, in individual years within the period, often represents 20% – 25% of topics. Topics related to news, another important genre of content during most of the 19th century, represent 17% of total topics, and often represent 20% of topics for individual years. Every issue had a section called Intelligence that was dedicated to news from the field. Additionally, individual articles, particularly in the earlier years of the journal, would be dedicated to translating articles published abroad and commenting on them and also on publishing letters to the editor that would discuss scientific endeavors both in the U.S. and abroad.
The topic models also demonstrate some other interesting, though not particularly surprising trends. Above is a simple line graph showing the number of topics within particular categories; the graph shows that geology topics increase over time, whereas other topics generally decrease. The graph also shows that until about 1871, “other sciences” were actually significantly higher than geology. Also in 1871 “other sciences” decline precipitously and geological topics increase and overtake “other sciences.” Since the American Journal of Science is currently a journal dedicated to geology, one would expect to see this trend. It is interesting to note, however, that this shift happens in the period from 1871 to 1897. The 1890s are a period when multiple other scientific professional societies are created, along with related scholarly journals. For instance the Journal of the American Chemical Society was founded in 1879 and the American Physical Review (journal for the American Physical Society, the society for physicists) began in 1893. The trend line for chemistry topics also shows a decline during this period. Clearly more detailed analysis of these topical trends is needed. Nonetheless, the trends illustrated in this line graph may be evidence of scientists leaving the more generalized American Journal of Science for more specialized journals when they are created. The decline of “other sciences” does seem to happen at exactly the right period of time.
Finally, I have one more graph that shows much the same data, however it represents the topics as a percentage of the whole, rather than as raw numbers of topics as shown in the line graph. This graph of percentages presents some nuance to the picture presented in the line graph. Geology topics represent fewer than 30% of the entire number of topics in 1819, and that number gradually increases to nearly 40% in 1922. Conversely, other sciences represent a high of nearly 60% in 1845, but decrease to a low of about 35% in 1922. Thus, one can see that other sciences are still an important number of topics even as late as 1922. This could complicate the story about scientists departing to other journals. It is possible that many scientists, despite the appearance of alternative journals, are still choosing to publish in the American Journal of Science. Additionally this relatively high percentage of “other science” topics could simply demonstrate that geology is a discipline that requires knowledge of other disciplines such as physics or biology in order to perform geological work. Again, more research and closer reading of the individual articles represented by these rather broad topics is needed to better understand how individual scientists are responding to a changing scholarly communication landscape.
The gradual decline of other sciences in these graphs may demonstrate that the nature of the authority within the American Journal of Science changed over time. As other societies created their own authority in competing journals like the Journal of the American Chemical Society, scientists within fields such as physics and chemistry decided to publish their work in those other venues. At the same time, many scientists, particularly geologists, continued to publish in the journal long after the death of Benjamin Silliman, the journal’s founder, in 1864. Therefore, one has to assume that the journal created a kind of authority that outlasted its founder. The nature of that authority, most likely, is through the same kinds of trust-building that other journals established, such as affiliation with a professional scientific society, peer-review, and reliance on authors’ credentialing within university hiring systems. Perhaps the method of topic modelling and text analysis by itself cannot answer the question of how authority is constructed. Topic modelling can, however, provide a useful source of evidence that identifies trends for further investigation and can be used to further strengthen traditional historical analyses of the history of scholarly communication.