Contacts between classes repeated more often than in fixed degree graphs over five days at one French school

We tested repeated contacts between teaching classes in a school record from 2 to 6 December 2013. The main sample had 326 adults with consistent class labels. For four successive day transitions, we compared observed repeats with simple bipartite graphs that fixed each student's daily number of partners in each other class. Observed repeat rates were 11.4% to 15.8%. Reference means were 6.8% to 9.3%. Observed/reference ratios were 1.68 to 2.11. All four observed counts exceeded 4000 saved reference draws. The approximate Holm-adjusted tail estimate was 0.001. A separate control fixed degrees within each hour and took the union across hours. Its ratios were 1.23 to 1.53, with observed counts above each 95% reference range. That control had more daily distinct edges, so it does not isolate a schedule effect. Longer sampling runs and both source class assignments gave the same direction. A 300-second contact threshold gave weak evidence on the last transition. The package includes code, aggregate outputs, figures, source hashes, and retrieval instructions. These results show repeat partner structure beyond the stated degree controls in one record. They do not establish friendship, a causal memory effect, or disease spread.

contact networks · repeated contacts · school classes · fixed degree graphs · temporal networks · SocioPatterns · reproducible analysis