The geometric tail distribution is characterized as the only distribution for which the regression of the record spacing of size two: Rn2 – Rn on the record Rn is constant. Up to now the only related problems considered in the literature were concerned with adjacent records. The proof of the result is based on a solution of a difference equation which unexpectedly was much harder to deal with, than an analogous equation which arises if the same problem for weak records is considered – see Wesolowski and Ahsanullah (2001).