[Author Login]
Title:Second language data and constraints on Manner: explaining substitutions for the English interdentals
Authors:Linda Lombardi
Abstract:Second language data and constraints on Manner: substitutions for

the English interdentals

Linda Lombardi

University of Maryland, College Park

The English interdental fricatives are marked, uncommon sounds.

For speakers of English as a second language (L2), two different

substitutions are seen - [t] or [s] (ignoring voicing).

Interestingly, speakers of a given first language (L1) tend to

all use the same substitution: e.g., speakers of Thai and

Russian use [t] while speakers of Japanese and German use [s].

It is difficult to explain such substitutions in a rule-based

theory. L1 has no [theta], so the child could not have

acquired a rule changing it either to [t] or to [s] - that is,

this cannot be an obvious case of what the L2 acquisition

literature calls transfer. Yet, some aspect of L1 must be

crucial, since speakers of different languages use different

substitutions. As I will show, Optimality Theory (OT) allows a

solution to this conundrum. While these speakers cannot possibly

have learned a rule that applies to [theta], they will have

independently necessary constraint rankings in their grammar that

will inevitably have some effect on [theta]. I will further

argue that these data provide support for particular

formalizations for markedness and faithfulness constraints on

Manner features.

First I show that stops are less marked than fricatives, so

that the ranking *cont >> *stop is universal. This is supported

by sound system generalizations, and by first language

acquisition data: all children go through a stage where

fricatives are absent and are replaced by the corresponding

stops. I will then propose that the initial state of UG contains

a unitary constraint on faithfulness to the Manner class, Ident

Manner, but that there may be reranking of the component

constraints on individual Manner features on the strength of

positive evidence. We can then account for the different

replacements as follows:

1. [theta] -> [t] Markedness high ranked: Output is a stop,

the less marked Manner

2. [theta] -> [s] Faithfulness high ranked: Output is more

faithful to the Manner of the input

The grammar that accounts for (1) contains the original ranking

of constraints on Manner that the first language learner begins

with. In an L1 whose speakers show this substitution, this

original ranking has not changed. In contrast, where L2 speakers

show (2) there must be some L1 phonology that forced reranking

of the constraints on Manner, which we should be able to

identify. I will exemplify these predictions with data from

Japanese, Thai, Dutch and Italian.

In sum, we can interpret (2), then, as a transfer effect -

something particular to that L1 results in that substitution -

and (1) as an effect of universals - the UG ranking, retained in

this L1, causes this substitution. But crucially, both

substitutions are simply the result of applying the L1 grammar

(constraint ranking) to the L2 data: no L2-specific mechanisms

are necessary.

In addition, this paper shows that L2 data can shed light on

properties of UG that are relevant to L1 phonology, in this case

the formulation of the markedness and faithfulness constraints

that apply to Manner features.
Type:Paper/tech report
Article:Version 1