Next-day contact recurrence exceeded a random model with fixed student partner counts by class in one school week

We tested whether contacts repeat after daily partner counts and class structure are held fixed. The data cover 327 observed students in nine classes during five days in December 2013. A daily pair needed at least 20 seconds of total recorded contact. Both students had to have a raw contact on the next day. Across four transitions, 3,825 of 8,202 eligible pair opportunities repeated: 46.63%. A random model with fixed class totals gave 29.42%. A model that also fixed each student's partner count to each class gave 36.32%. Its central 95% simulation range was 35.70% to 36.95%. The added constraints reduced excess recurrence by 40.10%. Each transition retained an excess. Excluding the partial first day retained a 9.95 percentage point excess. At a 300-second cutoff, observed recurrence was 41.05%, compared with 11.79% in the stronger model. Independent counts, exact small-graph checks, longer chains, and a seeded replay supported the calculation. This is one school week. Model ranges do not measure population uncertainty. The residual can include seating, schedules, and small groups. It does not establish friendship or a causal memory effect. The package preserves code, source hashes, model draws, figures, and aggregate tables. Raw data remain external under source terms.

contact recurrence · temporal networks · school contacts · conditional random model · degree constraints · SocioPatterns