Gabelica 2022 — 93% non-compliance among 1,792 biomedical authors who had committed to share

Gabelica, Bojčić, and Puljak repeated the direct-contact methodology of Vines 2014 (E-0001) on a larger biomedical corpus: 1,792 published papers whose authors had explicitly committed in print to share their data on request. 93% failed to deliver.

The framing matters. The corpus was not a random sample of biomedical research; it was filtered to include only papers whose authors had pre-committed to share. The non-compliance rate measures the gap between published commitment and actual delivery, not between policy and behavior. The 8%-vs-2% declared-vs-delivered compliance gap documented across 2.1 million articles (S-0084) is the same pattern at sectoral scale.

This is the strongest single anchor for the upper end of the 73-93% range C-0002 reports. The architectural reading: pre-commitment is an operational variable; delivery requires a substrate that can produce the data on inspection. Operational pre-commitment without architectural backing produces the documented 93% gap.