Unveiling citation peaks: How innovation faces delayed recognition in science

Abstract

Biases persist in the recognition of scientific innovation. To address the limited quantitative exploration, we conducted a systematic theoretical and empirical investigation. Integrating Innovation Diffusion Theory and Academic Capital Theory, we established hypotheses on citation dynamics. Utilizing over 24 million pre-2010 publications with at least 10 citations from the Microsoft Academic Graph dataset, we developed a kernel density estimation (KDE)-based method to model citation peaks in annual citation series. Using synthetic citation datasets with diverse patterns as benchmarks, we demonstrated that our KDE-based method consistently outperforms traditional peak detection methods in both peak identification accuracy and feature capture. Correlations between the first citation peak lags we identified and existing Sleeping Beauty indices validated our model. We find that over 31 % of observed publications exhibit multiple citation peaks, and their later peaks typically originate from more distant disciplines compared to earlier peaks. By introducing dual metrics—the novelty and disruption indices—we assessed innovation levels and found that delayed recognition of innovative research is pervasive and consistent across research fields, team sizes, h-index quartiles, and publication venues. Compared to conventional papers, the most innovative papers tend to reach their citation peak 1–2 years later and have a 20 % higher likelihood of exhibiting multiple peaks. Their highest peaks generally occur later, and they exhibit peak durations approximately 0.3 years longer with a more even citation distribution. Regressions with interaction terms indicate that larger or established teams and high-impact journals typically accelerate recognition, especially for conventional research, while their effect moderately diminishes for highly innovative work. Our study enhances the theoretical and methodological foundations of innovation recognition and provides insights to foster innovation dissemination and equitable research evaluation.