72 Rebecca Roeder
contrast, and they give the categorical example of vowel lengthening in Eng-
lish before voiced consonants (122) —an adjacency effect that is not purely the
result of physical limitations in the human voice mechanism. Stevens (1998:
288) also brieÀ y discusses the variation in overall vowel patterns that occurs
across regional dialects, but the degree to which edge effects vary based on
dialect has not been fully investigated.
The current study addresses the question of whether adjacent phonetic
environment—speci¿ cally consonantal context—displays the same patterns
in Mexican American speakers from a Northern Cities Shifted (NCS) area
as were found in earlier studies—namely Hillenbrand, Clark and Nearey
(2001) and Stevens and House (1963). Although the participants in the
Hillenbrand et al. study were also NCS speakers, the results did not conÀ ict
signi¿ cantly with the results of the Stevens and House study. Therefore,
any variation found in the Mexican American speakers may indicate either
- unique characteristics attributable to the inÀ uence of Spanish or 2) dis-
tinctive characteristics of NCS English which emerged because the elicited
speech was more natural.
Stevens and House (1963) measured formant frequency values for three
men pronouncing the eight vowels /i, Ԍ, ͑, NJ, ľ, ࣜ, ࡱ, u/ in the following three
environments: 1) in isolation, 2) in /hVd/ syllables, and 3) in stressed, sym-
metrical CVC syllables, preceded by unstressed /hԥ/ (e.g., /hԥbV b /, / hԥdVd/,
etc.) with the 14 consonants /p, t, k, b, d, g, f, v, s, z, ș, į, ߿, ˎ/. S u r p r i s -
ingly, no signi¿ cant F1 or F2 differences were found between the ¿ rst two
environments—vowels in isolation and in /hVd/ syllables; therefore, these
two sets of tokens were combined into a category that was somewhat con-
fusingly labeled the null context (Stevens and House 1963: 116) for purposes
of comparison to other environments. ReÀ ecting the contemporary perspec-
tive of the ¿ eld, no additional demographic information was offered on the
three informants.
The over-arching result of the Stevens and House study was that conso-
nantal environment (excluding the /hVd/ environment) causes vowel central-
ization, or undershoot. A number of more speci¿ c systematic effects, some of
which are discussed further below, were also found to correlate with manner,
voicing, and place of articulation features of the adjacent consonants.
Hillenbrand et al. (2001) expanded on Stevens and House (1963) in several
ways. They interviewed six men and six women, all but one of whom were
from the same general Northern Cities Shifted dialect area. In order to main-
tain comparability with the earlier study, they chose the same eight vowels,
and used a subset of the consonants (initial /h, b, d, g, p, t, k/ and ¿ nal /b, d,
g, p, t, k/). The vowels were recorded in isolation, and in CVC syllables read