The evidence given would seem to allow the strictest possible reading of the universal:IF there are word-initial geminates, THEN these same geminates will also occur word-medially.Muller mentions lgs where word-initial geminates are limited to subsets of word-medial geminates.Continuants, esp. nasals, seem the preferred segment class to occur as geminates in any position, and esp. word-initially. (Yapese only has /l, g/ as geminates, but none of the nasals.)
The evidence given would seem to allow the strictest possible reading of the universal:IF there are word-initial geminates, THEN these same geminates will also occur word-medially.Muller mentions lgs where word-initial geminates are limited to subsets of word-medial geminates.Continuants, esp. nasals, seem the preferred segment class to occur as geminates in any position, and esp. word-initially. (Yapese only has /l, g/ as geminates, but none of the nasals.)