Time-sliced Scientometric Analysis — Tracking Science Across Time Periods

There is no universal rule. Five- to ten-year windows are standard in most published studies because they balance temporal resolution against statistical stability. For rapidly evolving fields (machine learning, COVID-19 research) three-year or even annual slices may be appropriate. For slow-moving humanities disciplines, ten-year windows are common. The key requirement is that each slice contains enough records (typically at least 50–100 papers) for indicators to be meaningful. The window length must be justified in the paper and sensitivity analyses with alternative windows are good practice.

The terms are largely synonymous in practice. Longitudinal bibliometric study is the broader label; time-sliced scientometric analysis is a more specific operationalization that emphasises the explicit partitioning of the corpus into discrete, non-overlapping periods before computing indicators. Both contrast with cross-sectional bibliometrics, which treats the entire corpus as a single aggregate.

VOSviewer and bibliometrix (R package) are the most widely used tools and both support period-based analysis. CiteSpace specifically implements time-sliced analysis as a core feature, generating burst detection and cluster maps per period. The choice depends on the indicators needed: bibliometrix excels at statistical indicator tables; VOSviewer at network visualisation; CiteSpace at detecting temporal bursts and research fronts.

Yes, and this is increasingly recommended. The scientometric layer (publication trends, keyword evolution, collaboration networks by period) provides a quantitative map of the field's structure and dynamics. A complementary qualitative reading of key papers from each period then interprets the meaning of the structural changes. This mixed approach — sometimes called bibliometric-enhanced systematic review — adds depth that pure counting cannot provide.

Citation lag is a genuine problem: papers published in the final one to three years of the study window have had less time to accumulate citations and will appear artificially less influential. Common strategies include: noting the lag explicitly in the limitations section, excluding the final two years from citation-based comparisons while including them in publication-count analyses, or using normalised citation indicators that correct for document age.