What L1 and L2 acquisition tell us about the phonological faculty and general cognition

This paper surveys a range of core phonological effects in first- and second-language acquisition:
 
(i) spontaneous emergence of opacity in acquisition of L1 (Jesney 2005) and L2 (Idsardi 2002), derived environment effects (Kiparsky and Menn 1987, Eckman and Iverson 1995), and unnatural processes (Smith 1973, Bedore et al. 1994, Edwards 1996, Gierut and Champion 2000); 
(ii) inter- and intra-grammar variation (Tropf 1987, Vihman and Greenlee 1987,Vaux 2005); 
(iii) convention/hypergeneralisation (abstraction of rules in the absence of unambiguous evidence; Tenenbaum 1999, Vaux 2006) and the 3-exposures-to generalisation rule (LouAnn Gerken,p.c.); 
(iv) interlanguage (Vaux 2005); 
(v) avoidance (Celce-Murcia 1977, Kellerman 1977);
 
and investigates their implications for phonological theory. Specifically, the surveyed effects are shown to be: 

a. incompatible with parallelist models such as Optimality Theory and Connectionism and their attendant learning theories (the Gradual Learning Algorithm (Boersma and Hayes 2001), Recursive Constraint Demotion (Tesar & Smolensky 2000), and error-driven learning (Gluck 1991 inter alia)); 
b. consistent with a theory in which phonological generalisations are captured in the form of ordered, deterministic, symbolic rules abstracted from the primary linguistic data by general cognitive mechanisms that employ information theoretic learning (Gallistel 2003) and strong-sampling Bayesian inference (Tenenbaum & Griffiths 2001).