On the quest for currencies of science. Field “exchange rates” for citations and Mendeley readership
| DOI | https://doi.org/10.1108/AJIM-01-2017-0023 |
| Published date | 18 September 2017 |
| Date | 18 September 2017 |
| Pages | 557-575 |
| Author | Rodrigo Costas,Antonio Perianes-Rodríguez,Javier Ruiz-Castillo |
| Subject Matter | Library & information science,Information behaviour & retrieval,Information & knowledge management,Information management & governance,Information management |
On the quest for currencies
of science
Field “exchange rates”for citations
and Mendeley readership
Rodrigo Costas
Centre for Science and Technology Studies (CWTS), Leiden University,
Leiden, The Netherlands and
Centre for Research on Evaluation, Science and Technology (CREST),
Stellenbosch University, Stellenbosch, South Africa
Antonio Perianes-Rodríguez
SCImago Research Group, Department of Library Science and Documentation,
Universidad Carlos III de Madrid, Getafe, Spain, and
Javier Ruiz-Castillo
Department of Economics, Universidad Carlos III de Madrid, Getafe, Spain
Abstract
Purpose –The introduction of “altmetrics”as new tools to analyze scientific impact within the reward
system of science has challenged the hegemony of citations as the predominant source for measuring
scientific impact. Mendeley readership has been identified as one of the most important altmetric
sources, with several features that are similar to citations. The purpose of this paper is to perform an in-depth
analysis of the differences and similarities between the distributions of Mendeley readership and citations
across fields.
Design/methodology/approach –The authors analyze two issues by using in each case a common
analytical framework for both metrics: the shape of the distributions of readership and citations, and the field
normalization problem generated by differences in citation and readership practices across fields. In the first
issue the authors use the characteristic scores and scales method, and in the second the measurement
framework introduced in Crespo et al. (2013).
Findings –There are three main results. First, the citations and Mendeley readership distributions exhibit a
strikingly similar degree of skewness in all fields. Second, the results on “exchange rates (ERs)”for Mendeley
readership empirically supports the possibility of comparing readership counts across fields, as well as the
field normalization of readership distributions using ERs as normalization factors. Third, field normalization
using field mean readerships as normalization factors leads to comparably good results.
Originality/value –These findings open up challenging new questions, particularly regarding the
possibility of obtaining conflicting results from field normalized citation and Mendeley readership indicators;
this suggests the need for better determining the role of the two metrics in capturing scientific recognition.
Keywords Citation analysis, Scientometrics, Altmetrics, Currencies of science, Mendeley readership,
Reward system of science
Paper type Research paper
Introduction
In 1998, Garfield stated that “[t]he Mertonian description of normal science describes
citations as the currency of science. Scientists make payments, in the form of citations, to
their preceptors.”The idea of citations as the currency of science was also discussed one
year later by Wouters (1999), according to whom “the role of the citation might also be
compared with that of money, especially if the evaluative use of scientometrics is taken into
account. Whenever the value of an article is expressed in its citation frequency, the citation
is probably the most important unit of a ‘currency of science’.”Thus, citations are often seen
as a means for rewarding scientists for their work and scientific merit, so that, together with
Aslib Journal of Information
Management
Vol. 69 No. 5, 2017
pp. 557-575
© Emerald PublishingLimited
2050-3806
DOI 10.1108/AJIM-01-2017-0023
Received 12 January 2017
Revised 27 April 2017
18 May 2017
29 May 2017
Accepted 29 May 2017
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/2050-3806.htm
557
On the quest
for currencies
of science
authorship and acknowledgments, they have become an integral part of the so-called
“reward triangle”(Cronin and Weaver, 1995).
This role of citations as the main currency in evaluative scientometrics has gone
unchallenged for many years. The recent emergence of new ways of measuring the
reception of scientific publications by different audiences in the form of the so-called
“altmetrics”(Haustein et al., 2015; Priem et al., 2010) probably represents the most important
attempt at expanding the system of scientific currencies. This development of new
indicators, aimed at capturing broader perspectives of the symbolic capital of scientists’
outputs, may cause a change in the “rules”and “norms”of a more multifaceted reward
system of science, where different forms of symbolic capital might interplay in the
representation of the esteem and recognition of scientific agents (Desrochers et al., 2015).
However,in spite of the initial expectationsregarding altmetrics as potentialalternatives to
citations (Prie m et al., 2010), recent research on the most important social media metrics
(e.g. Twitter, Facebook, blogs, etc.) suggests that there are fundamental differences with
citations: in coverage (Thelwall et al., 2013), main characteristics (Haustein et al., 2015),
correlations (Costas et al., 2015b; Haustein et al., 2014), and interpretation within the scientific
reward system (Haustein et al., 2016). These results essentiallyhighlight the limited potential
of most of these metrics as realistic alternatives to citations. Consequently,their role has been
relegated to complementary metrics capturing aspects that are not covered by citations
(Cronin, 2013; Torres-Salinas et al., 2013; Costas et al., 2015a, b; Haustein et al., 2014).
There is however an altmetric source that has been highlighted as an exception to this
pattern: Mendeley readership. The Mendeley database, identified as one the most important
sources of alternative metrics (Wouters and Costas, 2012;Torres-Salinas et al., 2013),has been
found to have a high coverage of scientific publications (i.e. a large shareof publications have
some readership in Mendeley, cf. Zahedi et al., 2014); there are moderate correlations between
readership and citations (Zahedi et al., 2014; Li et al.,2012), and readership scores also have a
good filtering ability of highly cited publications (Zahedi et al., 2017). Field differences in the
presence of Mendeley readership (Costas et al., 2015a), and the technical possibilities of
calculating field normalized Mendeley readership indicators (Fairclough and Thelwall, 2015;
Bornmann and Haunschild, 2016; Haunschild and Bornmann,2016) have also been discussed
in the literature. Consequently, it becomeshighly relevant to study the characteristics of field
readership distributions and to test the feasibility of alternative fieldnormalization strategies
with the same techniques used for studying and normalizing field citation distributions.
Objectives
This paper has three objectives. First, to study whether field Mendeley readership
distributions for a large set of Web of Science (WoS) publications are as highly skewed and
as similar across fields as is found for field citation distributions. Second, to explore the
possibility of overcoming the field dependence of Mendeley readership practices by
estimating “exchange rates (ERs)”for comparing Mendeley readership counts across fields
as has been done for comparing citation counts. Third, to compare the consequences for field
normalization of using Mendeley ERs and mean readership as normalization factors.
Methodology
Data and analytical approach
We consider a total of 1,125,811 distinct publications labeled as “articles”in the WoS from
year 2012, with a DOI and belonging to any of the 30 broad fields used by Ruiz-Castillo and
Costas (2014)[1]. The problem of the assignment of publications to more than one field
category is solved following a multiplicative approach (cf. Herranz and Ruiz-Castillo, 2012).
The corresponding extended count in the final data set consists of 1,634,932 records, whose
558
AJIM
69,5
To continue reading
Request your trial