Replication Rate by Citation Count
Are highly cited papers more replicable? Yang, Youyou & Uzzi (PNAS 2020) found that the 61 papers that failed replication were cited at the same yearly rate as the 39 that passed — citation counts did not distinguish replicable from non-replicable findings (their Fig. 1). This page repeats that comparison on the full replications database, using per-year citation counts from OpenAlex for every original paper with a DOI.
Rate = success / (success + failure + reversal); inconclusive and unrecorded outcomes excluded, where success means the outcome recorded in the database. Unit: replication effect (n = 7,382 — originals matched in the OpenAlex citation dataset). How this is defined.
Citation trajectories for papers that passed vs failed replication
Median cumulative number of citations received by each year since publication, for original papers classified as replicated (2,376 papers) vs not replicated (1,869 papers) under the selected criterion and threshold. Unlike Yang et al.’s single-year 2008 cohort, our originals span several decades, so the x-axis is years since publication rather than calendar year; each point uses only the papers old enough to have observed that year, and the axis is capped at 30 years (or earlier if a group falls below 30 papers) — beyond that, the shrinking sample is dominated by a few heavily-replicated classics.Shaded bands are distribution-free 95% confidence intervals for the median (from order statistics); where they overlap, the two groups’ citation rates are statistically indistinguishable. Citation counts are heavily right-skewed — the mean runs roughly 3× the median because replication projects target influential papers — so the median is the default view. Hover the lines for exact values and group sizes.
4,245 papers in the trajectory chart · 999 excluded (no OpenAlex match or publication year) · 899 excluded (inconclusive outcome)
Replication rate by citations in first 5 years
6,084 replications plotted in the binned charts · 1,298 determinate rows excluded (no OpenAlex match or fewer than 5 observed years)
Citation source. Total citation counts and per-year trajectories come from the OpenAlex API (generated 2026-07-13). A paper’s per-year citations are the publication years of the works citing it; the current partial year is excluded from trajectories, and the rare citing works dated before publication (preprints, metadata errors) are counted in year 0.
Matching. Original papers are matched to OpenAlex by DOI: 7,398 of 8,892 replication rows (83.2%) have a matched original; rows without a DOI or without an OpenAlex record are excluded from the charts.
Units. The trajectory chart classifies each original paper as replicated if at least the threshold share of its determinate effect-level replications succeeded. The binned charts instead count every determinate replication attempt once. Reversals count as determinate non-replications; inconclusive rows are excluded.
Data: replications_database_2026_09_28_155521.csv.