🚀 go-pugleaf

******************** LingCircle - Wednesday 3/31 *********************

Papers to be presented by Hartwell Francis and Bill Raymond

3:30 - 5:00 pm
Old Geology, Rm 114
Refreshments provided

Hartwell Francis - ARE LEXICAL SUBJECT DEVIANT?

Bill Raymond - EVIDENCE FOR THE ANALOGICAL BASIS OF LANGUAGE VARIATION:
                 ENGLISH ARTICLES IN PRODUCTION AND PERCEPTION
 

                               ABSTRACTS
*************************************************************************

Hartwell Francis - Are lexical subject deviant?
===============================================

A examination of the tagged portion of the Switchboard corpus reveals that
subjects are mostly pronominal.  This paper focuses on the 9% of the
subjects that are lexical.  Through the use of lexical subjects
speakers appear to violate the Principle of Seperation of Reference and
Role.  However, we find that speakers who use lexical subjects appear to
constrain hearer processing load by ensuring that these lexical subjects
denote referents which are accessible or anchored to the discourse
context.  This claim is based on two morphosyntactic trends in the coding
of lexical subjects: (1) the use of definite determination and (2) the use
of pronominal possessives and relative clause modifiers.



Bill Raymond - EVIDENCE FOR THE ANALOGICAL BASIS OF LANGUAGE VARIATION:
                   ENGLISH ARTICLES IN PRODUCTION AND PERCEPTION
======================================================================
                William D. Raymond, Julia A. Fisher,
                Alice F. Healy, and Lyle E. Bourne, Jr.

        Morphological forms often exhibit allomorphic variation
conditioned by grammatical factors of the context in which they appear. In
rule-based grammars variation is captured using inviolable, categorical
rules that apply deterministically. However, variation in performance is
often not strictly rule-like. Alternative explanations of variation employ
violable constraints (Optimality Theory, Prince & Smolensky 1993), or
violable, probabilistic rules (Variable Rule models, Cedergren & Sankoff
1974). These models offer greater descriptive precision, but have been
criticized as lacking explanatory adequacy (Guy 1992). One possible
explanatory mechanism is analogy. Recent theories of analogy have made
explicit the factors to which analogical variation is sensitive (Bybee
1995), and connectionist models can replicate analogical behavior as a
consequence of network learning (Hare & Elman 1995). To this point,
analogical explanations have been limited to phenomena of language
acquisition (Rumelhart & McClelland 1986) and language change (Hare &
Elman 1995).

        A test case for exploring an analogical account of synchronic
performance variation is definite and indefinite article variation in
English. Article variant use is largely determined by phonological
environments: distinct forms are used preceding words beginning with
consonants and vowels for the indefinite article, a or an, and, in most
dialects, the definite article "the" (pronounced /D/ or /Di/). However,
language behavior with respect to article use deviates from this rule, as
shown by corpus analysis (Todaka 1992) and experimentation (Healy et al.
1997). Research has suggested that there are a number of speaker factors
to be taken into account, including planning issues (Fox Tree and Clark
1997), dialect, and age (Keating et al. 1994). Much less is known about
lexically determined variation, which is the focus of this study.

For this study, two experiments probe the extent to which article
variation is influenced by lexical factors by manipulating the
CONSONANT-VOWEL CONTEXT, but also WORD FREQUENCY, LEXICAL STRESS,
PHONEME-TO-GRAPHEME PAIRINGS, INITIAL VOWEL QUALITY, and GRAMMATICAL
CATEGORY of the following word. We examined participants preferences in
wholly auditory tasks involving production and perception. In Experiment
1, 24 participants repeated words combined with their choice of variant in
either a definite or an indefinite article condition. In Experiment 2, for
the same two conditions, another 24 participants heard the same words with
the variants and chose which of the two they preferred. Results are as
follows:

1 CONSONANT-VOWEL CONTEXT explains about 75% of the variation with the
definite article but more than 97% with the indefinite article.

2 Analyses of variance show a significant effect of varying
PHONEME-TO-GRAPHEME CORRESPONDENCE, with more errors on words with
non-canonical pairings ("united") than canonical ones (yellow) even though
participants saw no written materials.

3 SYLLABLE STRESS interacts with CONSONANT-VOWEL CONTEXT, reflecting the
fact that segment  type is significant only for first syllable stress
words.

        Differences between behavior on definite and indefinite article
variants is explained as a result of the relationship between indefinite
article forms and the explicit orthographic differences in indefinite
article variants. Both factors limit possibilities for analogical
variation. However, definite article variants display a pattern of
variation analogous with lexical variation of reduced vowels as a function
of syllable onset and stress. The pattern is consistent with factors
identified as defining analogical processes by Bybee 1995. The effects of
phoneme-to-grapheme pairing on both definite and indefinite article
variation corroborate research showing effects of orthography on language
use independent of reading, but also underscore the significance for
behavior of conflicts between representational types.