| |
Abstract:
This paper demonstrates the existence of long-domain
coarticulatory patterns associated with English /l/ and /r/,
describing the extent and nature of differences in articulation
and acoustics. Most work on coarticulation to date has focussed
on the coarticulatory effects of vowels on adjacent segments or
across consonants. The coarticulatory effects of consonants have
been relatively understudied, despite the claim that consonants
such as /l/ and /r/ exert long-distance coarticulatory effects,
or "resonances". In fact, consonantal coarticulatory effects (C
to V) are generally acknowledged as smaller, with a more
restricted temporal range than V to C effects. This paper shows
that the coarticulatory effects of /l/ and /r/ are strikingly
long-ranging, extending up to two syllables in the anticipatory
and perseverative directions. Simultaneous EMA and EPG recordings
of three speakers of standard Southern British English were made.
6 minimal pairs of real CVC words, with initial consonant either
/l/ or /r/, were produced in the frame sentence "Have you uttered
a __ at home?". General Linear Models, supplemented by t-tests
where necessary, were constructed for the acoustic and
articulatory data (F1, F2, F3 measurements and EMA x,y coil
positions) collected at the temporal midpoint of vowels of
interest. Consonantal data (EMA and EPG) were collected at the
point of maximum closure in the EPG signal (all consonants
examined were alveolar stops). The acoustic and articulatory data
showed retracted and/or raised tongue position, lip rounding and
F3 lowering in the /r/ relative to the /l/ context, up to two
syllables remote from the liquid, for all speakers. Some
individual variation in temporal extent was apparent, but the
overall picture is one of surprising consistency: l/r
coarticulatory effects are distributed across several syllables,
with all 3 speakers showing anticipatory coarticulation two
syllables before the liquid. These data pose problems for most
recent theories of coarticulation, which do not attempt to model
such long-distance effects. In particular, the school of
Articulatory Phonology and its associated dynamic gestural
modelling considers coarticulation to be the result of overlap or
blending of gestures. To explore the extension of gestural
overlap required to model these long-distance coarticulatory
effects, a model of gestures as the result of damped second order
differential equations is being implemented. EMA tongue movement
traces have been accurately modelled (to within the reported
measurement error of the EMA system), and the adjustments needed
to model /l/ vs. /r/ tokens are currently being examined.
|
