Short abstract:
 
 The same categorical phenomena which are attributed to hard
 grammatical constraints in some languages continue to show up as
 statistical preferences in other languages.  This is the "stochastic
 generalization", already observed in syntax by Givon (1979: 26-31),
 In this talk I will give an overview of recent evidence for the stochastic
 generalization in syntax and its unifying analysis within the
 framework of Stochastic Optimality Theory.  The example domains
 include -

   A. Cross-linguistic variation (passives in English and Lummi)
   B. Diachronic change (differential object marking in Spanish)
   C. Style shift (case ellipsis in Japanese and Korean)
   D. Language internal variation ("'s" and "of" genitives in English)