I read this interesting article by Letchford et al. (2015) on the correlation between paper title length and citation rate. The authors found that for the 20,000 most cited articles each year (from 2007-2013) — not discriminating by discipline — shorter papers are cited more often.
I wondered if this was true within one discipline and within one journal. Fortunately Geophysical Research Letters (GRL) has published some 30,000 articles since its inception in 1974. I downloaded the data for the 20,000 most cited GRL articles in 2015 from Scopus and compared the number of characters in each article’s title to the number of citations the article garnered in 2015:
- Note the typical skewed distribution of citations on the y-axis marginal
- Note the almost gaussian distribution of article title lengths on the x-axis marginal
- Thanks to @lukaszpiwek for his work Tufte in R
Ok, so is there a trend? I followed some nice instructions on R tutor to determine if GRL title length (
TL) and the number of citations received in 2015 (
C) were correlated using Kendall’s tau, a nonparametric correlation metric that was used in Letchford et al., (2015).
> cor(GRL2015, method="kendall", use="pairwise") Title.length X2015.citation Title.length 1.0000000000 0.0007179821 X2015.citation 0.0007179821 1.0000000000
Tau, a measure of the correlation that varies form 1 to -1, is very low (0.0007), so I tested to see if the null hypothesis (i.e., title length and citations are uncorrelated) can be rejected:
>cor.test(TL, C, method="kendall") Kendall's rank correlation tau data: TL and C z = 0.14206, p-value = 0.887 alternative hypothesis: true tau is not equal to 0 sample estimates: tau 0.0007179821
So we cannot reject the null hypothesis that Title Length and 2015 citation counts are uncorrelated.
There has been other work on this subject, with mixed outcomes. Take a look at the works cited in Letchford et al., (2015) if you are interested…