The Initial and Final State: Theoretical Implications and 
 

Empirical Explorations of Richness of the Base 







Paul Smolensky, Lisa Davidson, and Peter Jusczyk



Johns Hopkins University











One of the more elusive aspects of Optimality Theory, the Richness of
 

the Base principle asserts that all regular language-particular
 

patterns are imposed by the grammar on its outputs, with no
 

corresponding structure present in the inputs to the grammar. A
 

theoretical consequence for the final state of the grammar of a
 

language lacking X is that markedness constraints M prohibiting X must
 

outrank faithfulness constraints F demanding preservation of X in the
 

input/output mapping. Frequently no explicit evidence for this final
 

ranking is provided by the grammar, since there is often no need to
 

posit X in any underlying forms. Thus a further consequence is that if
 

such a final state is to be learnable, the ranking M >> F must be
 

imposed by the learning process, and already be present in the initial
 

state of the grammar.  In this paper we present empirical paradigms we
 

are developing to assess these predictions of Richness of the Base.
 

	INITIAL STATE. Infants from 4.5 to 20 months of age were
 

presented triples of syllables of the form 'A B AB' in which 'AB' was
 

either a faithful concatenation of A and B, or one in which a
 

markedness-reducing sound change had occurred. Under the hypothesis
 

that infants prefer stimuli which conform to their grammar, and the
 

interpretation /A + B/ -> [AB], the prediction of Richness of the Base
 

is that sound-change stimuli should be prefered over faithful but
 

marked stimuli. This was confirmed by the Headturn Preference Procedure
 

for children at 4.5, 10 and 20 months, although no significant
 

difference was found at 15 months.



	FINAL STATE. Adult English speakers' productions of non-English
 

onset clusters were elicited. We sought an experimental paradigm that
 

would induce speakers to subject non-native inputs to their English
 

grammars, to quantitatively assess the predictions of Richness of the Base
 

that non-English clusters would be repaired to meet the requirements of
 

English syllable structure. In the condition that best approximated this
 

prediction, non-English clusters divided into several groups which could
 

be ordered according to their probability of repair. Appropriate
 

interaction of independently motivated constraints can account for the
 

relative markedness of different non-English clusters, suggesting a final
 

English ranking that makes such distinctions, without apparent motivation
 

in the English data in which none of these clusters appear.  Possible
 

analyses of how such a ranking might arise are suggested. Such 'hidden
 

rankings' in the final state are important for pursuing the hypothesis
 

that the initial state for L2 acquisition is the final state for L1.