Web links as analogues of citations by Alastair G. Smith

Garfield, Eugene garfield at CODEX.CIS.UPENN.EDU
Mon Jul 26 11:36:09 EDT 2004

Information Research, Vol. 9 No. 4, July 2004


Web links as analogues of citations
Alastair G. Smith
School of Information Management
Victoria University of Wellington
New Zealand

This exploratory study investigates the extent to which Web links are
analogues to the citations in traditional print literature. A classification
of Web links is developed, using the nature of the source and target pages,
and the reasons for linking. Links to a sample of research oriented Websites
(universities, professional institutes, research institutes, electronic
journals, and individual researchers) were classified. Overall, 20% of the
Web links in the study could be regarded as research links analagous to
Citations in conventional print publications have traditionally been used as
indicators of links between researchers, and it is tempting to regard Web
links as analogues of citations. To what extent do hypertext links between
Websites indicate research links? This exploratory study examines the nature
of Web links made to research oriented Websites, to determine the extent to
which the Web links are made for research-related reasons and are,
therefore, analagous to citations.

One obvious difference is the nature of the documents that are linked.
Citations in conventional print publications are generally between research
publications, while Web links may be between a wide variety of publication
types: personal home pages, subject resource guides, etc.

Why are citations and Web links made? Bibliometric literature contains many
studies that examine the motivation behind citing. Egghe and Rousseau (1990)
review a number of studies of citation motivation. Garfield (1964)
identifies reasons that include paying homage, identifying methodology,
providing background, correcting, criticising, substantiating or
authenticating, alerting to forthcoming or poorly disseminated work,
identifying original concepts, and disclaiming or disputing previous work.
Kelland and Young (1998) list several motivations, including acknowledging
prior work, identifying methodology, providing background reading,
correcting or criticising, substantiating claims, alerting readers to
forthcoming work, authenticating data, identifying original publication of a
term or concept, disclaiming work of others, and disputing priority claims.
Case and Higgins (2000) found that citations might be made because the cited
work (1) was a 'concept-marker', (2) promoted the authority of the citing
work or (3) deserved criticism. Oppenheim and Smith (2001) studied citations
in students' dissertations and found that while their motivations for citing
were similar to academics, there was a trend towards citing Internet

Studies of linking motivation on the Web, on the other hand, are more
recent, and indicate that Web links are made for different reasons. Kim
(2000) studied links from Web based scholarly publications and found a wide
range of motivations for linking, some echoing citations, but also
including: publicity, credit to an institution, providing an immediate
access mechanism, to provide a graphical image, and an editorial policy of
encouraging hyperlinks. Thelwall (2003) studied 100 Web links between
university home pages, and classified them into 'ownership', 'social',
'general navigational' and 'gratuitous'; arguing that the majority of Web
linking motivations were trivial compared with citing motivation. Chu (2003)
analysed links to academic Websites, and found that 50% of links were to
resource or directory information, while only 27% were motivated by research
or teaching/learning. Because of these wider reasons for Web linking, and
the more eclectic range of documents that Web links are made between,
motivations are different for Web links and citations. Wilkinson et al.
(2003) found that less than 1% of links to university department Websites
were formal research citations. Vaughan and Shaw (2003) studied
bibliographic and Web citations to articles in LIS journals and found that
many Web citations represented intellectual impact and that for most
journals there was a correlation between Web citations and the Journal
Impact Factor.

The increasing use of Web links in Webmetric research means that the nature
of these links, and the extent to which they serve the same function as
citations, is important. The current study undertook to classify links made
to a sample of research oriented Websites, to test a trial classification
and to estimate the extent to which the Web links were analagous to

Fifteen research-oriented sites were studied, taking three examples of each
of five types of sites. These were:

University Websites (Victoria University of Wellington , Australian National
University, and MIT)
Professional institute Websites (The Royal Society of New Zealand, Institute
of Professional Engineers of New Zealand, and the New Zealand Law Society )
Research institute Websites (National Institute of Water and Atmospheric
Research, Institute of Veterinary, Animal and Biomedical Sciences, and the
New Zealand Institute for Economic Research)
Electronic Journals (British Medical Journal , Journal of Internet Law and
Technology , and AtP (AntePodium))
Individual researchers' Web pages (an NZ artificial intelligence researcher,
a US library and information management researcher, and a US mathematical
These sites were chosen as being representative of a range of ways in which
research appears on the Web, and representing both academic and professional
research. This approach to selection was appropriate for an exploratory
study testing the classification and methodology.

Links to each of the target sites were determined by using the AltaVista
command in advanced mode (http://altavista.com/Web/adv):

link:xxx and not host:xxxwhere xxx is the target domain. This excluded Web
links made within the site itself. Site collapse was off, and the search was
set to be world wide, for documents in English (to avoid having to classify
sites in languages the researcher was not familiar with). Every twentieth
item on the list was examined, up to a total of ten linking sites, except
where fewer than 200 linking sites were found, when every tenth site was
examined. If a site or link was no longer valid, the next link on the
results list was chosen. 150 links were studied in total. Searches were
carried out in January 2003. As this was an exploratory study,
classification was carried out by the researcher without independent
verification of the classifications.

This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email

More information about the SIGMETRICS mailing list