[Author Login]
Title:The inadequacy of Filters and Faithfulness in Loanword Adaptation
Authors:Carole Paradis
Abstract:The objective of this paper is twofold. First, it aims to show that segment

deletion in borrowings is largely predictable, and that this predictability

might be problematic for a filter-based framework since, as we will see, it

entails that phonological processes are visible to phonological constraints.

For instance, it is at odds with a filter-based framework such as Optimality

Theory (OT; Prince & Smolensky 1993 and McCarthy & Prince 1993), where

constraints are in fact "filters", i.e. surface constraints, which do not

have access to the processes of the phonological component or the intermediate

forms they generate since, by definition, filters deal with final outputs only.

Second, this paper will demonstrate that ill-formed segments contained in

borrowings are adapted, i.e. recast into a different shape (85.2% of cases),

or left unadapted (10.7% of cases) - if these are imports - instead of being

deleted. Phonologically-induced segment deletion represents only 2.3% of cases,

a fact which is attributed to a principle of the Theory of Constraints and

Repair Strategies (TCRS; Paradis 1988a,b), the Preservation Principle. It will

be shown that, although the Preservation Principle and Faithfulness (Parse and

Fill) in OT seem comparable, they make indeed different predictions. In the

view of Faithfulness, the optimal output (candidate) is the one which has

undergone the least (segment) deletion, i.e. the one whose segments are all

"parsed" (realised), in addition to being the one which has undergone the

least (segment) epenthesis (Fill). In other words, the Faithfulness constraints

(Parse and Fill) establish that the best candidate is the one which is as close

as possible to its input. OT treats both Faithfulness constraints on a par in

the sense that their ranking with respect to one another - as is the case with

any other constraint in this framework - is determined on language-specific

grounds, not universal ones. This means that, statistically, segment deletion

in loanwords across languages should be observed as often as segment insertion,

a prediction which is not empirically supported.
Type:Paper/tech report
Article:Version 1