Citation Chaining: Find More Research From One Paper

A software engineering student is writing about technical debt: the hidden cost of quick fixes that make code harder to change later. She has found one excellent recent paper on the topic.

That single paper is more useful than it looks. Its reference list points back to the work it built on, all the way to Ward Cunningham’s 1992 experience report, where the “debt” metaphor first appeared. And the papers that have cited it since point forward to the latest research.

Following those links, backward and forward, is called citation chaining. It is one of the oldest search methods in research, and still one of the most effective.

What is citation chaining?

Citation chaining is a way of finding research by following the citations of a paper you already know is relevant. Backward chaining looks at the sources a paper cites. Forward chaining looks at the newer papers that cite it.

You will also see it called citation tracking, citation searching, reference mining or snowballing. The idea is the same in every case: let researchers’ own citations lead you to related work.

Backward vs forward citation chaining

	Backward chaining	Forward chaining
Direction	Into the past	Toward the present
What you look at	The paper’s reference list	Newer papers that cite the paper
What it finds	Foundational work, earlier studies, the origins of an idea	Follow-up studies, critiques, updates, applications
Where to do it	In the paper itself	In a citation index, such as Google Scholar’s “Cited by” link, Scopus or Web of Science
Main limit	Can only go back to what the authors chose to cite	Recent papers have had little time to be cited

For the technical debt paper, backward chaining leads to the classic sources: Cunningham’s original metaphor and the early studies that defined the field. Forward chaining leads to newer work on measuring technical debt, managing it in agile teams and detecting it automatically.

Used together, they show both where an idea came from and where it is going.

Why citation chaining works

The best evidence comes from researchers who audited their own searches.

In 2005, Trisha Greenhalgh and Richard Peacock looked back at a large review they had carried out on how organizations adopt innovations, drawing on 13 different research traditions. They traced where each of the review’s 495 sources had come from:

Only 30% came from the planned search: database searches and hand-searching 32 journals.
51% came from snowballing, mainly following references of references.
24% came from what the team already knew or from colleagues they asked.

The effort behind those numbers is just as striking. The database searching took about two weeks of a specialist librarian’s time, and produced one useful paper for every 40 minutes of searching. Citation tracking of selected key papers took around a day in total, and found three reviews that no other method had turned up.

Their conclusion was that reviews of broad, complex topics cannot rely on database searches alone, however many databases are searched.

Software engineering researchers have reached a similar view. In 2014, Claes Wohlin of the Blekinge Institute of Technology in Sweden published guidelines for snowballing in systematic literature studies. He tested them by repeating a published review and concluded that snowballing, used as the first search strategy, may be a good alternative to database searches.

The reason is simple. Researchers cite the work that matters to their argument, whatever words it uses and wherever it was published. A citation is a link that a keyword search can’t see.