Twitter predicts citations of ecology research

As scientists, communicating our research is just as important as doing it. We are well-trained to use the scientific method—we make hypotheses, we conduct experiments, draw conclusions, and repeat. But if nobody knows about what we did, what’s the point?

That’s the main reason The Fisheries Blog exists. We use a popular medium to communicate fisheries science. But after nearly five years as a team, we started to think more broadly. Because most of us on the team are professional researchers, we wanted to know if there was a link between the amount of social media activity about an article and its research impact.

So, we put on our research hats and our #scicomm hats, and wrote a paper that came out last week in the journal PLoS ONE. Read the full article for free here:

We asked a simple question: Does Twitter activity predict citation rates of ecological research? We focused on Twitter instead of Facebook or blogs because it is home to a thriving forum of scientists communication with one another. We chose ecology as a discipline, rather than fisheries because:

  • there are many more general ecology journals than fisheries journals
  • there are likely many more ecologists than there are fisheries scientists on Twitter due to the size of the discipline
  • fisheries is a subset of ecology

We used a mixed-effects model (with journal identity as a random effect) to estimate effects of Twitter on citation rates (from the Web of Science database). Our data came from nearly 1,600 journal articles published from 2012-2014 in 20 ecology journals. We collected several altmetrics (alternative citation metrics) that quantify Twitter use:

  • number of tweets about an article
  • the number of users that sent those tweets
  • twitter reach–how many people saw those tweets

We also needed to control for the effects of other factors that can affect citation rates, namely:

We found several key results:

  • Number of tweets had significant positive effects on the number of citations
  • The most important predictor of citation rates in our dataset was time since publication
  • 5-year journal impact factor had a strong, but imprecise effect on citation rates
Relative importance (vertical) and effect sizes (horizontal) of parameters in generalized linear mixed models predicting citation rates of 1,599 primary research articles in ecology.

Relative importance (vertical) and effect sizes (horizontal) of parameters in generalized linear mixed models predicting citation rates of 1,599 primary research articles in ecology. Error bars represent 95% confidence intervals

Based on our results, it was possible that the effect of Twitter was masked by journal impact factor. Were articles from higher-impact journals more discussed (and thus more cited) on Twitter? We used simple correlations and more complex mixed effects models to get at the question, and the answer was a solid no. In other words, the effect of Twitter use was generally independent of impact factor.

Ours is certainly not the first paper to look at the relationship between social media activity and citation rates (see our full paper for a discussion). However, ours is unique in that we account for social media activity, time since publication, and journal impact factor in the same model.

Several studies have found weak bivariate correlation between citations and various altmetrics, and have concluded that altmetrics capture something altogether different than traditional citations. Our results suggests that the true relationship may be masked by other factors (mainly time since publication). Accordingly, we see the value in assessing research impact using both traditinoal citations and altmetrics. We see it as a conversation of ‘and‘, rather than ‘or‘.

Within the scope of the articles we examined, our results suggest that good research will be discussed on Twitter and cited based on its merit and relevance to scientists in niche subdisciplines, regardless of journal impact factor.  This doesn’t mean that papers in high-impact journals won’t get cited (obviously they do), just that the effect of impact factor is more variable.

Model-predicted effects of two variables that significantly affected citation rates

Model-predicted effects of two variables that significantly affected citation rates

Conversely, it’s doubtful that you can ‘game the system’ to get more citations by over-tweeting your research. Number of tweets was highly correlated to number of users, suggesting that the power of Twitter is more in the collective, rather than single influential users tweeting about a paper (although that is certainly possible). There’s no shortcut to a well-cited paper.

So, scientists, should you be on Twitter promoting your research? On the one hand, the relationships we found are clear. On the other hand, going on Twitter and doing nothing but touting your own research isn’t likely to gain you any friends, followers, or citations. I enjoy Twitter because it is forum for lively and current scientific discussion, and I appreciate that it is rarely a bunch of scientists talking ‘at’ one another. I’ve discovered several highly relevant papers on Twitter that my Google Scholar and ResearchGate alerts did not pick up. Occasionally, a cool paper gets discussed; we’ll see what happens to this one.


Please leave a thought provoking reply. We reserve the right to remove comments deemed inappropriate.

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s