[Author Login]
Title:Syllable types in cross-linguistic and developmental grammars
Authors:Claartje Levelt, Ruben van de Vijver
Comment:19pp. Word and RTF files require SILDoulosIPA font (only for five characters)
Abstract:Syllable types in cross-linguistic and developmental grammars

Claartje Levelt & Ruben van de Vijver

In Optimality Theory (OT) (Prince & Smolensky 1993) the basic

assumption is that constraints are universal, but that the rankings

of these constraints are language specific. For language typology the

idea is that different rankings reflect grammars of different possible

languages, while for acquisition the idea is that the learner needs to

acquire the specific ranking of the mother-tongue.

Languages are cross-linguistically marked to a greater or lesser extent

with respect to structural constraints that refer to syllable type

(ONSET, NO-CODA, *COMPLEX-ONSET, *COMPLEX-CODA). The assumption here is

that the grammar of a totally unmarked language is identical to the

initial grammar (G1) of the language learner (Gnanadesikan, 1995). When

the language to be learned is marked in several respects, the hypothesis

is that the learner acquires these marked aspects by gradually promoting

Faithfulness above the Structural constraints in the hierarchy. There

is, thus, a learning path where the learner, going from G1 to the final

grammar (Gf), passes through several intermediate grammars. The

hypothesis is that the intermediate grammars in development are also

final state grammars of languages of the world. Vice versa, it is

expected that, since languages can be marked in different respects,

there are different possible learning paths which the learner can take

to reach the final state.

This hypothesis is tested here, by combining data on syllable types

in different languages from Blevins (1995) ("cross-linguistic grammars")

with data on the acquisition of syllable types (developmental grammars)

of children acquiring Dutch (Levelt, Schiller & Levelt 1997).

It turns out that cross-linguistic grammars differ from the

developmental grammars in two respects. First, some intermediate

developmental grammars apparently do not correspond to any cross-

linguistic grammar. Some stages require a grammar where local

conjunctions of constraints play a role (Smolensky 1993). The question

is whether cross-linguistically a grammar which makes use of such local

conjunctions exists, or whether local conjunctions in developmental

grammars are acquisition-specific and reflect, for example, transient

processing problems.

Second, there is less variation in the intermediate grammars than could

be expected on the basis of the cross-linguistic data. The second

intermediate grammar, G2, corresponds to the grammar of Thargari, not

to the grammar of Cayuvava, which is logically equally possible, G4 is

Mokilese, not Sedang or Klamath. The only variation is for G5: Finnish

for one group of learners, Spanish for another group of learners. The

explanation given here is that in case alternative developmental steps

are theoretically possible, information from the input pushes the

learner in a particular direction. Syllable type frequencies in both

child- and adult-directed Dutch speech exactly predict the choices for

a specific intermediate grammar, and also predict where variation will

Type:Paper/tech report
Article:Version 1