Alerted by a post on Nature News and Comment, I read with interest a newly posted paper from Mikko Packalen and Jay Bhattacharya from the National Bureau of Economic Analysis entitled "Age and the Trying Out of New Ideas."
The abstract of this working paper states:
Older scientists are often seen as less open to new ideas than younger scientists. We put this assertion to an empirical test. Using a measure of new ideas derived from the text of nearly all biomedical scientific articles published since 1946, we compare the tendency of younger and older researchers to try out new ideas in their work. We find that papers published in biomedicine by younger researchers are more likely to build on new ideas. Collaboration with a more experienced researcher matters as well. Papers with a young first author and a more experienced last author are more likely to try out newer ideas than papers published by other team configurations. Given the crucial role that the trying out of new ideas plays in the advancement of science, our results buttress the importance of funding scientific work by young researchers. (Emphasis added)
Needless to say, I was intrigued. After a quick read, I looked deeper into the methodology, particularly with regard to the highlighted terms above.
The study is based on the use of MEDLINE (accessed through PubMed). More precisely, they used “Author-ity” MEDLINE, a previously constructed version of MEDLINE with the names of authors disambiguated as much as possible. This database was used for two purposes. First, new ideas were identified by searching titles and abstracts for two- or three-word strings and associating these with the year when they first appears. Strings that subsequently occurred with high frequency were deemed to be important new ideas. The Nature commentary includes a list of the ten most frequent concepts for each decade and inspection reveals these to be sensible. Second, the "age" of each investigator was estimated by determining the year in which the first publication by this investigator appeared. Thus, this is "career age" rather than chronological age. This is a sensible approach which has both advantages and disadvantages. Most importantly, it is workable from the available data. I know from some of my recent analyses, estimating chronological ages of investigators can be quite difficult. In addition, this automatically at least partially corrects for increasing training periods over time. A disadvantage is that an early publication can "age" an investigator compared to peers.
With these two parameters, the authors were set to do some analysis. Figure 1 in the paper is shown below:
Panel A shows the fraction of publications trying out new ideas versus the career age of the first author. Clearly, there is a broad peak in which the first authors are within 3-12 years of his/her first publication. This, of course, primarily reflects the accomplishments of graduate students and postdocs!
Panel B shows the comparison for All Authors. This shows a more featureless downward trend. This probably reflects the contributions of graduate students and postdocs but with more senior members of the research teams in the mix.
Panel C shows the distribution with Key (both first and last) Authors. This shows a peak in the career age range of 10-15 years. Given the first author distribution from Panel A, this suggests that the last author distribution has a peak around 20 years. This is confirmed by the data presented in Figure 3 which shows the data in two-dimensional format with Career Age of Last Author versus Career Age of First Author.
The maximum for this post occurs with first authors with career ages between 1 and 10 years (i.e. graduate students and postdocs) and last authors with career ages between 8 and 25 years (early to mid-career faculty).
Does this mean that early to mid-career investigators are the most productive in trying out new ideas? Yes and no. As a population, they certain do appear to be. However, these data have not been normalized (as far as I can tell) to the distribution of the ages of investigators. This distribution (which has been changing over time as we were recently reminder by Drugmonkey) shows peaks (both for R01 grantees and medical school faculty) in the range of 45-55 years old over the period covered by this analysis. If one assumes an age of the beginning on independent careers of 36 (over this period not just at present), the data are consistent with the number of faculty at each career stage being an important factor.
Overall, the paper clearly supports the roles of graduate students and postdocs in being first authors of many (most) papers that appear to break new ground. This is, as George Carlin would say, "near-fetched." The results regarding last authors will require a more careful reading of the paper and, perhaps, more analysis by the authors. However, this is clearly an important data set and approach to provide empirical evidence that bears on these important issues.