At fixed counts of 2 to 10 messages in week one, two-way CollegeMsg pairs had an uncertain contact difference over the next 28 days

We tested whether contact in both directions in the first seven days was linked to later contact. We used 6,489 first-observed CollegeMsg pairs with 2 to 10 early messages. Both direction groups were standardized to the same exact early count distribution. Contact in the next 28 days occurred at a rate of 13.83% for two-way pairs and 12.29% for one-way pairs. The difference was 1.54 percentage points. Central 95% descriptive ranges were -0.39 to +3.48 points with pair resampling and -2.20 to +5.00 points with shared-person weights. These rules do not establish a population confidence claim. Adding early message timing gave +0.39 points. Requiring contact on two later days gave +0.11 points. A post-plan calendar adjustment gave +0.17 points. These checks can change the overlap sample. The result does not establish a later contact advantage or prove that the difference is zero. The record is one online community in 2004. Absence of later messages does not establish tie loss. A separate scan checked every primary pair feature. The canonical download matched the retained mirror. A full repeat matched all result files. The package has code, aggregate data, source hashes, resampling draws, figures, and repeat instructions. Raw source records remain external under their original rights.

human communication · reciprocity · contact persistence · landmark analysis · CollegeMsg · null result