SMS direction changes within five minutes exceed an hourly shuffle reference in four weeks of Copenhagen student data

We tested whether local direction counts account for fast changes in SMS direction. We used the public four-week Copenhagen Networks Study SMS file. Among 14,642 eligible transitions in 565 participant pairs, 72.39 percent changed direction within five minutes. A uniform shuffle within each pair and hour kept exact pair event times and counts in each direction. Its exact mean was 54.81 percent. The excess was 17.59 percentage points. None of 9,999 shuffled records reached the observed rate. The one-sided Monte Carlo probability estimate was 0.0001 under this model. The hourly excess was 20.95 percent smaller than the excess under a shuffle over the full record. The excess stayed positive in all four weeks, with equal pair weights, and after removal of the most active pair or participant. Direction changes and temporal reciprocity are established findings. This work adds a local count audit and a reproducible benchmark. It does not identify replies or prove a causal memory process. The result concerns one student cohort and observed messages only. The package contains source data and provenance, code, exact means, null draws, validation checks, and figures.

temporal networks · SMS · reciprocity · conditional randomization · human communication · Copenhagen Networks Study