Editors’ note: Today’s post is by Marvin Starominski-Uehara, an Adjunct Assistant Professor at Temple University Japan.

There is a longstanding tradition in the sociology of science of treating self-citation with suspicion. When researchers reference their own previous work, the assumption is that this behavior inflates citation metrics without adding genuine intellectual value. Bibliometric tools used in hiring, promotion, and grant allocation often discount self-citations on precisely this reasoning.

But this suspicion has never been without critics.

In a recent study published in the journal Society, I examine self-referencing patterns across the most-cited publications in an emerging interdisciplinary field: human stigmergy. The main finding, that self-references are a statistically significant predictor of citations from unaffiliated authors while non-self-references predict nothing, is revealing but should be interpreted carefully.

The Argument To Make

It is tempting to frame empirical findings like this as a refutation of the conventional view. But the conventional view, at its most honest, is not that all self-citation is fraudulent. Barnes, et al. (2023), “In Defense of Endogeny,” for instance, argue that while higher rates of endogeny can help indexes identify journals being used for self-promotion, nepotism, or other unethical ends, endogeny itself should not be equated with them. Their case study illustrates how pioneering scholars are, almost by necessity, among the first contributors of a new journal.

What the Data Shows, and What It Cannot

The human stigmergy sample draws from 1,289 sources across nine languages, narrowed to 42 papers that met selection criteria. For each, I recorded self-references, non-self-references, and total citations received from authors with no connection to the original work.

The regression found that each additional self-reference was associated with approximately 11 additional citations from unaffiliated authors. The relationship held after removing the authors’ own subsequent self-citations from the total, which rules out the most obvious confound.

Non-self-references showed no meaningful relationship with citation outcomes. Authors who cited extensively from outside their own body of work were no more likely to attract return citations than those who did not.

These are interesting findings, but the regression explains only 19.3 percent of the variance in citation outcomes, meaning roughly 80 percent is driven by other factors entirely. Drawing on Pierre-Paul Grassé’s theory of stigmergy, I frame these factors as forms of peripheral stimulation: journal prestige, author reputation, institutional affiliation, research topic trends, and social networks. These environmental signals drive citation behavior with a force that individual referencing choices cannot match.

This matters for how we read the self-citation effect. Researchers who have accumulated an extensive body of prior work to cite are also, typically, researchers who have been in a field long enough, and published in sufficiently visible venues, to attract citations for reasons unrelated to their referencing habits. Self-citation may be functioning here less as a cause of impact than as a marker of the kind of cumulative engagement with a research program that tends to attract attention over time.

Termites and the Logic of Intellectual Construction

The theoretical framework behind this research draws on Grassé’s 1959 stigmergy model, originally developed to explain how termite colonies build elaborate structures without centralized coordination. Individual workers respond autonomously to modifications already made to the structure around them, each act of construction becoming a stimulus for the next.

The parallel is direct. When researchers cite their own prior work, they are responding to their own intellectual constructions and building on frameworks robust enough for others to engage with. The worker termite, Grassé observed, returns to the same site several times before moving elsewhere. Researchers who sustain engagement with their own intellectual territory are doing something structurally similar.

Descriptive Patterns Worth Noting

Beyond the regression, the descriptive statistics offer some context. Across the ten most-cited papers in the sample, self-reference rates ranged from 1.1 to 41.1 percent. Papers at the higher end tended to show longer citation lifespans, whereas those at the lower end often displayed concentrated citation bursts followed by sharper decline.

Foundational theoretical papers also appeared to benefit more from self-referencing than applied studies did. This is intuitive since researchers building comprehensive theoretical frameworks naturally accumulate more prior work to reference while applied studies tend to sustain relevance through continued real-world utility rather than intellectual lineage.

What This Means, What It Does Not

The reflexive discounting of self-citation in bibliometric evaluation may be penalizing exactly the kind of original, impactful work through which knowledge structures become durable enough for others to build on. That is the strongest claim this study supports.

None of this, however, is an argument for unlimited self-referencing. Excessive self-citation has been used to game metrics, and the literature on self-promotion through journals and citation cartels documents this clearly. The point is that the response to that concern, in the form of blanket bibliometric penalties, has been blunt in ways that create collateral damage for legitimate scholarship.

Emerging fields, niche specialisms and researchers engaged in the authentic work of building coherent theoretical programs tend to show higher self-citation rates for reasons that have nothing to do with gaming anything. The evidence from human stigmergy adds to the case that this pattern can correlate with genuine scholarly impact. In this field, at least, it was not how widely researchers engaged with others’ work that predicted lasting impact but how consistently they engaged with their own, and that is a pattern the literature on endogeny and field formation would have predicted all along.