ABSTRACT A STUDY OF THE PIN/PEN MERGER IN OKLAHOMA … · A STUDY OF THE PIN/PEN MERGER IN OKLAHOMA...
Transcript of ABSTRACT A STUDY OF THE PIN/PEN MERGER IN OKLAHOMA … · A STUDY OF THE PIN/PEN MERGER IN OKLAHOMA...
1
ABSTRACT
A STUDY OF THE PIN/PEN MERGER IN OKLAHOMA
By
Phillip Weirich
Oklahoma, being situated geographically and culturally between the Midlands and the South,
and only having seen widespread Anglo settlement as early as the late 19th century, presents an
interesting laboratory for the study of language change when different dialects come into
sustained contact. One particularly interesting linguistic phenomenon that follows such contact
is the merger of two previously distinct vowels. An earlier study of English in Oklahoma, SOD
(Bailey et al., 1997), identified a number of demographic characteristics associated with the
PIN/PEN merger, including age, gender, and location (urban vs. rural). Although results from
SOD painted a clear picture of the distribution of this merger (or for many speakers, a split) and
tracked its progress in apparent time, no mention was made about variation in production
within the merged and split speakers.
This study seeks to provide a description of the production of several speakers from a
range of relevant demographic groups in and around Oklahoma City. Wordlist production of
minimal and near-minimal pairs of 24 speakers from eight demographic categories (age,
urban/rural, gender) is considered. An impressionistic evaluation of speakers’ merger status
and acoustic analysis of formant contours are used to describe the sorts of variation present in
Oklahoman production of vowels in the PIN- and PEN-classes. The findings show that in
addition to having either merged or split PIN/PEN vowels, speakers also produce the vowels as
monophthongs, lax triphthongs, or tense triphthongs. All of the young speakers produced the
vowels as monophthongs, although some were merged and others were split. However, most of
the middle-aged speakers produced triphthongs, suggesting that Oklahoma City and
surrounding areas are shifting from a triphthongal norm to a monophthongal norm for these
vowels.
2
Table of Contents
1 INTRODUCTION ............................................................................................................................................ 6 1.1 GOALS .............................................................................................................................................................................. 6 1.2 ORGANIZATION OF SECTIONS ...................................................................................................................................... 7
2 BACKGROUND AND REVIEW OF RELEVANT LITERATURE ............................................................. 8 2.1 VOWELS: MERGERS AND SPLITS ................................................................................................................................. 8 2.1.1 Basic Characteristics of Vowel Systems ....................................................................................................... 8 2.1.2 Mergers ...................................................................................................................................................................... 9 2.1.3 Splits ......................................................................................................................................................................... 12 2.1.4 Near-‐Mergers ....................................................................................................................................................... 12
2.2 SOCIAL FACTORS OF CHANGE: DIFFUSION ............................................................................................................. 13 2.3 THE PIN/PEN MERGER .............................................................................................................................................. 14 2.3.1 Linguistic Factor: Nasalization .................................................................................................................... 15 2.3.2 Social Factor: Path of Diffusion .................................................................................................................... 16
2.4 OKLAHOMA: CULTURE AND LANGUAGE ................................................................................................................. 17 2.4.1 Settlement History .............................................................................................................................................. 17 2.4.2 Where is Oklahoma? .......................................................................................................................................... 20 2.4.3 PIN/PEN in Oklahoma ......................................................................................................................................... 23
3 METHODOLOGY: DATA COLLECTION .................................................................................................. 25 3.1 PARTICIPANTS ............................................................................................................................................................. 26 3.2 DATA COLLECTION ..................................................................................................................................................... 27 3.2.1 Interview ................................................................................................................................................................. 27 3.2.2 Wordlist .................................................................................................................................................................. 28
3.3 RECORDING AND PREPARATION FOR ANALYSIS ................................................................................................... 31 4 DATA ANALYSIS AND DISCUSSION ........................................................................................................ 32 4.1 IMPRESSIONISTIC EVALUATION ............................................................................................................................... 33 4.1.1 Methods ................................................................................................................................................................... 33 4.1.2 Results ...................................................................................................................................................................... 34
4.2 ACOUSTIC ANALYSIS .................................................................................................................................................. 36 4.2.1 Nasalized Vowels ................................................................................................................................................ 36 4.2.2 Types of formant contours .............................................................................................................................. 37 4.2.3 Monophthongs: “flat” countours .................................................................................................................. 38 4.2.4 Triphthongs: “humped” contours ................................................................................................................ 41 4.2.5 Diphthongs: “sloped” contours ...................................................................................................................... 43 4.2.6 Examples of intraspeaker variation ........................................................................................................... 45
4.3 SUMMARY OF LINGUISTIC FINDINGS ....................................................................................................................... 62 4.4 SOCIAL DISTRIBUTION ............................................................................................................................................... 65 4.4.1 Single Demographic Variables and Production Style ......................................................................... 65 4.4.2 Merged vs. Split and Tense vs. Lax Across Demographic Groups ................................................... 68
5 CONCLUSION ................................................................................................................................................ 70
3
LIST OF TABLES Table 2.1: State of origin of settlers of 1889 Land Run .......................................................... 20
Table 3.1: PIN/PEN vowels elicited in wordlist .............................................................................. 29
Table 3.2: hVd words presented in the wordlist ........................................................................... 30
Table 4.1: Results of Impressionistic Evaluation ......................................................................... 34
Table 4.2: Summary of individual speakers' production styles .................................................... 63
4
LIST OF FIGURES Figure 2.3: A schematized depiction of merger-by-expansion ..................................................... 12
Figure 2.4: The PIN/PEN merger in the North America ................................................................. 16
Figure 2.5: Map of Oklahoma and Indian Territories prior to 1889 ............................................. 18
Figure 2.7: Popular regions of North America ............................................................................. 21
Figure 4.1: Fitz "send" .................................................................................................................. 38
Figure 4.2: Shelly "sinned" ........................................................................................................... 39
Figure 4.3: Amber "tin" ................................................................................................................ 40
Figure 4.4: Dale "send" ................................................................................................................. 41
Figure 4.5: Allen "send" ............................................................................................................... 42
Figure 4.6: Sam "dim" .................................................................................................................. 43
Figure 4.7: Hilda "hem" ................................................................................................................ 44
Figure 4.8: Allen: Minimal and Near-Minimal Pairs .................................................................... 46
Figure 4.9: Fitz: Minimal and Near-Minimal Pairs ...................................................................... 48
Figure 4.10: Hilda: Minimal and Near-Minimal Pairs .................................................................. 50
Figure 4.11: Ramona: Minimal and Near-Minimal Pairs ............................................................. 52
Figure 4.12: Herbert: Minimal and Near-Minimal Pairs .............................................................. 54
Figure 4.13: Van: Minimal and Near-Minimal Pairs .................................................................... 56
Figure 4.14: Renee: Minimal and Near-Minimal Pairs ................................................................ 58
Figure 4.15: Amber: Minimal and Near-Minimal Pairs ............................................................... 60
Figure 4.16: Comparison of age and production style .................................................................. 66
Figure 4.17: Comparision of location of residence and production style ..................................... 67
Figure 4.18: Comparision of gender and production style ........................................................... 67
5
Figure 4.19: Comparison of demographic groups and merger status ........................................... 68
Figure 4.20: Comparison of monophthongal speakers' merger status .......................................... 69
Figure 4.21: Comparison of Triphthongal Speakers' usage of tense and lax vowel targets ......... 70
6
1 Introduction This thesis considers the phonological merger of PIN and PEN class vowels in and
around Oklahoma City, Oklahoma. The PIN/PEN merger, so called because speakers
pronounce the two vowel classes in the same manner and, thus, merge the two vowels
into one, has been present in American speech for over a hundred years (Brown 1990).
It is widespread in the South, and there are indications that the merger is expanding
beyond the traditional Southern dialect region (Labov et al. 2006). Oklahoma is a state
that was settled relatively recently by people from both Southern and Midwestern
states, so it provides an interesting laboratory in which to study dialectal change when
two dialects come into contact. An earlier study of the PIN/PEN merger in Oklahoma,
the Survey of Oklahoma Dialects (SOD) (see Bailey et al. 1997), found that the PIN/PEN
merger had nearly gone to completion by the 1930s. However, sometime after the
1930s, the merger began to reverse, and some speakers began to produce a distinction
between the two vowel classes (Bailey et al. 1995). It is evident, though, that the diverse
settlement history of the state continues to affect the language of residents in interesting
ways. The present study examines recordings of 24 speakers from Oklahoma City and
adjacent rural areas collected in 2012 and 2013 in association with the most recent
attempt to explore language in Oklahoma, Research on Dialects of English in Oklahoma
(RODEO). In addition to an impressionistic evaluation of the merger status of the
speakers’ PIN/PEN vowels, acoustic data is also presented to provide a clearer picture of
how speakers merge or split (produce a distinction between) these vowels.
1.1 Goals When this project began, I thought that it would be a simple matter of recording several
speakers from various demographic groups, determining who merges and splits the
7
PIN/PEN vowels, and then looking at how mergers and splits are distributed across the
demographic groups. The previous literature on the topic from Oklahoma and other
linguistically similar states led me to believe that the two classes of vowels were either
produced the same way or not. As I began my analysis of the data I had collected, it
quickly became apparent that there was much more variation among speakers than
previously expected. Not only do speakers merge and split the vowels, they do so
using different vowel qualities.
Because no earlier studies in Oklahoma have considered the variation present in
the PIN and PEN vowels, my purpose is to describe the various means of production
used by speakers, and thereby add depth to a topic that has previously only considered
a binary description of speakers as being either merged or not merged.
The goals of this project are to describe 1) the ways in which speakers in the new
data set produce vowels in the PIN/PEN classes, 2) who among them merge and split the
vowels, and 3) how these identified features might provide insights into the future
direction of English in Oklahoma. These descriptions are made based on
impressionistic evaluation and acoustic analysis of minimal and near-minimal pairs
produced in a word list elicitation.
1.2 Organization of Sections Chapter 2 will consider relevant background information regarding the nature of
phonological change and vowel mergers as well as providing a summary of the
findings of an earlier Survey of Oklahoma Dialects (SOD). In Chapter 3, the
methodology of the study is presented, explaining how respondents were selected and
recorded and how the data was prepared for analysis. The results are presented and
discussed in Chapter 4. In this chapter, results of impressionistic and acoustic analysis
8
are presented as well as several examples from individual speakers for comparison.
Chapter 5 concludes the study by summarizing the relevant findings and suggesting
avenues for further study.
2 Background and Review of Relevant Literature This chapter presents relevant background information related to sound change, vowel
mergers in particular, to provide a context to understand why vowel mergers might
occur. Then it considers the cultural and linguistic history of Oklahoma including
previous research findings in the state regarding the PIN/PEN merger in order to suggest
why that particular vowel pair is undergoing change in Oklahoma.
2.1 Vowels: Mergers and Splits This section focuses on how vowel systems are organized and how they adapt to
change, especially as a result of mergers and splits.
2.1.1 Basic Characteristics of Vowel Systems Vowel systems, or any sound system for that matter, are actually quite resilient to
change, so it is usually somewhat of a surprise when a change arises. Two ideas that
originated in the 19th century remain helpful when thinking about sound change: the
desire for ease of articulation and the need for clarity. On the one hand, sounds need to be
easy to articulate in order to be efficient, but on the other hand, sounds must be distinct
enough that we can clearly distinguish one from the other.
This need for clarity has been explored through the principle of maximum
dispersion. Thomas (2011: 274) summarizes this rather old, but still relevant, idea in
9
saying that “contrastive sounds tend to become located as far apart as possible in the
perceptual space, as if they repel each other like the same pole of two magnets.” There
is evidence that some processes akin to maximum dispersion are at work in the world’s
languages, as with Liljencrants and Lindblom’s (1972) finding that the most common
three-vowel system was /i, a, u/ and not a less dispersed system such as /i, e, ɨ/.
Moulton (1962) even showed that vowels can “move” within the vowel space
depending on where the other vowels are located. For example, in his study of local
dialects of Swiss German he found that in dialects with an /æ:/ but no /ɔ:/, the /a:/
tended to shift backward, away from the /æ:/, while those dialects with an /ɔ:/ but no
/æ:/ tended to have an /a:/ that was more fronted.
Maddieson (1984) adds an important note about maximally dispersed vowel
systems. He points out that the most common vowel system is not actually maximally
dispersed because there are many ways to increase distinctions that are not used. For
example, we don't see systems that are /i, a, uˤ /. He says that “the world’s languages
only add the additional parameters of contrast to vowels if they include a fairly wide
sample of simple contrasts on the vowel quality dimensions” (16). So even though
maximal dispersion plays a role in determining the structure of vowel systems, other
factors, such as the desire for ease of articulation, are involved as well.
2.1.2 Mergers
Mergers are a particular type of sound change in which two sounds that were
historically distinct have become a single sound in both production and perception.
That is they result in a decrease in phonemic distinction. Vowel mergers seem unlikely
when considering the principle of maximum dispersion from the previous section
10
because vowels tend to avoid other vowels in order to maintain a distinction, as seen in
examples from Swiss German; however, various phonological and/or phonetic factors
can bring two vowels close enough together in the vowel space that a merger is
possible. The PIN/PEN merger is likely driven by the influence of the following nasal.
This possibility is discussed further in section 0.
Labov (1994: Ch 11) mentions two principles that vowel mergers tend to follow:
1) Garde’s Principle: Mergers cannot be reversed by linguistic means
2) Herzog’s Corollary: Mergers expand at the expense of distinctions
Merger’s cannot be reversed by linguistic means, according to Garde’s Principle,
because the two sounds have become identical; there is no longer a way for speakers to
distinguish the sounds and separate them into their original classes. Mergers can be
reversed by social factors, such as the influx to an area of large numbers of immigrants
who have a vowel system that do not have the merger (Bailey et al. 1993; Herold 1990).
Herzog’s Corollary says that mergers expand at the expense of distinctions,
meaning that if someone were to have to choose between a system that merges two
sound classes and a system that distinguishes two sound classes, the speaker is more
likely to choose the system with fewer sounds.
2.1.2.1 Three Kinds of Mergers
Three processes for the progression of mergers have been suggested. Merger-by-
approximation and merger-by-transfer were proposed by Trudgill and Foxcroft (1978)
while (Herold 1990; Herold 1997) furnishes the third, merger-by-expansion.
Merger-by-approximation (Figure 2.1) is the result of speakers producing the two
sounds with increasingly similar acoustic characteristics, which eventually converge at
some point in the acoustic space between the original vowels. If this kind of merger
11
applied to the PIN/PEN merger (which is not necessarily the case), the words “pin” and
“pen” would be pronounced in the same way with a new vowel that is neither as high
as “pin” nor as low as “pen.”
Merger-by-transfer (Figure 2.2), rather than being based on acoustic distinctions,
is accomplished by speakers recategorizing sounds from one phoneme and transferring
them to another nearby phoneme’s category. In other words, two vowel classes may
remain in the same position in the vowel space, but individual words begin being
pronounced with the vowel of another class. Using “pin” and “pen” as an example
again, [pin] might be used for both words, “ten” would be pronounced as [tin], etc.
j:€; a++T $:+2Ei
jEr?;eir ifsEi:ziis t:i:
ajSi€F:1 si+; ii;
tiji=i,ir;t18:L =:;i?i=.l'"iiiiiii tir:li;ijiT"e$jgE IS*;!tl:*TTit;:: u$;*gi:1i13s Yi€ j
H, ij li\i: EF itirEi ; ,=iE;; s*ii Siii=*Ei E !; \i:t:Peej:F€;+ g j iiE€i
iFd'to-do!a)ooLq)c.9 cl
o--tro.rydEc)s
a\O
i-
00)Lb0lt
A[/
00E-fl=i:
r,t€d
L,t-\
g/sa*
\ c
/-l' v\
c/.T*1 \ >
Fl-;;..,'rl
-i\ii! X +/ {
!\'_ ?.-
t _
\ iw.af /
cJ
vr/ fE'6=o
,/ iitr.\ ,/
\ 'E
/{+ rr\ / & \
.e/", {:sr\
i "
\ o-
l'zf ^l>
| 1>
C)I
'{l\\,ri"*/
\ '|
E\*+/
\-r / Y'
\__,, \___/
EE€/s\/,r{\/.1'\lr,l *+\
l* +\
Fr>
, lQ
o
li.?i -*l{ l.t"
- \
r \
- .' l
YV \u*,i
g.s0It
u(,zIUozf-a\8lr
I-vjs,E if !xr
iiiiiiiiaUUzrLo@+o"{
j:€; a++T $:+2Ei
jEr?;eir ifsEi:ziis t:i:
ajSi€F:1 si+; ii;
tiji=i,ir;t18:L =:;i?i=.l'"iiiiiii tir:li;ijiT"e$jgE IS*;!tl:*TTit;:: u$;*gi:1i13s Yi€ j
H, ij li\i: EF itirEi ; ,=iE;; s*ii Siii=*Ei E !; \i:t:Peej:F€;+ g j iiE€i
iFd'to-do!a)ooLq)c.9 cl
o--tro.rydEc)s
a\O
i-
00)Lb0lt
A[/
00E-fl=i:
r,t€d
L,t-\
g/sa*
\ c
/-l' v\
c/.T*1 \ >
Fl-;;..,'rl
-i\ii! X +/ {
!\'_ ?.-
t _
\ iw.af /
cJ
vr/ fE'6=o
,/ iitr.\ ,/
\ 'E
/{+ rr\ / & \
.e/", {:sr\
i "
\ o-
l'zf ^l>
| 1>
C)I
'{l\\,ri"*/
\ '|
E\*+/
\-r / Y'
\__,, \___/
EE€/s\/,r{\/.1'\lr,l *+\
l* +\
Fr>
, lQ
o
li.?i -*l{ l.t"
- \
r \
- .' l
YV \u*,i
g.s0It
u(,zIUozf-a\8lr
I-vjs,E if !xr
iiiiiiiiaUUzrLo@+o"{
Figure 2.1: A schematized depiction of merger-by-approximation (Thomas 2011: 285)
Figure 2.2: A schematized depiction of merger-by-transfer (Thomas 2011: 285)
12
Finally, merger-by-expansion (Figure 2.3) is the result of speakers broadening the
proportion of the vowel space used in the production of both vowels until there is no
longer a boundary between the two. In this case, “pin” might be produced as [pin] or
[pɛn], and the same for “pen.”
Figure 2.3: A schematized depiction of merger-by-expansion (Thomas 2011: 286)
2.1.3 Splits Splits are the opposite of mergers in that they result in an increase of phonemic
distinction rather than a reduction. According to Garde’s principle, mergers cannot be
reversed by linguistic means, but social factors are not excluded from playing a role in
reversing mergers and causing a split. For example, Britain (1997) reported the
expansion of an /u/ ~ /ʌ/ distinction among the Fens in Eastern England which was
apparently favored by social prestige (in Labov 2010: 335). In this case, the distinction
was borrowed from another dialect because of the perceived prestige associated with
the split vowel system and did not arise spontaneously within the Fen’s dialect.
2.1.4 Near-‐Mergers Near-mergers lie on a continuum between mergers and splits. Labov et al. (1991: 33)
say of near-mergers that “sound change may bring two phonemes into such close
approximation that semantic contrast between them is suspended for native speakers of
1\q,a{U(,zI(.)ozl
aiij*iiii;lllitiiiriiiigi:git
iiIiirlirieiijjiliiiiii*iiiiiii iiii igss:ij jg€ ig* +i;* lssi ?*ri;i
dtri--
qU-uzIo€oc{
; a-\ /:\Ev
I:-/
13
the dialect, without necessarily leading to merger.” Referring back to the depiction of
merger-by-approximation in Figure 2.1, we can imagine that time 2 represents a near-
merger in that the two phonemes are less distinct from each other than in time 1, but
enough of a distinction still exists that the two are not completely merged. Hearers may
not report being able to distinguish a minimal pair, even when two sounds are
measurably different. In production, too, the same people who cannot perceive a
difference may pronounce the minimal pair differently. For example, Labov et al. (1972)
reported about New York that “the native speaker hears source and sauce as ‘the same,’
but produced a reliable statistical difference between the nuclei of these vowels” (cited
in Labov 2006: 25). As long as some kind of distinction is maintained, regardless of if
speakers can consciously recognize the distinction, the sounds could eventually become
distinct in both production and perception.
2.2 Social Factors of Change: Diffusion
The geographic diffusion of linguistic innovations is perhaps the most readily
studied form of language change. Bailey et al. (1993) outline some of the more well-
attested types of diffusion. The most common sort of diffusion seems to be hierarchical,
in which “innovations begin in central places, which serve as the focal points for
diffusion across the landscape. Rather than spreading uniformly across the landscape,
diffusion begins in large cities like London, moving then to smaller cities and so on
down the hierarchy” (Bailey et al. 1993: 361). This does not appear to be the only kind
of geographic diffusion, though. Contrahierarchical diffusion has also been attested in
which features of local speech migrate into the population centers (Bailey et al. 1993;
Brown 1991). Bailey et al. also mention that linguistic diffusion isn’t as simple as
14
looking at geographic spread. Changes also occur along social lines, including ethnicity
and economic status (e.g. Labov’s department store study, 1966).
Prestige also plays an important role in language change and diffusion. Labov
(1972) refers to change from above and change from below. The latter concerns innovations
that begin within certain subgroups of a speech community and are “below the level of
social awareness” (Labov 1972: 178). These innovations gain traction as they become
associated with the subgroup and can acquire a level of covert prestige that serves a
function of solidarity within that group. Change from above, on the other hand,
concerns innovations that are imposed on speakers by some sort of linguistic authority
figure, such as parents, news broadcasters, teachers, and style manuals. Overt prestige is
associated with change from above because the speakers are usually aware of the
features and know that they are considered “correct” within the community. It has
been noted by many sociolinguists (Labov, 1966, for one example) that it is the
upwardly mobile middle class that are most concerned with overt prestige, while the
working class tend to harbor covert prestige.
2.3 The PIN/PEN Merger The PIN/PEN merger is a conditioned merger because the phonetic environment provides
the conditions necessary that lead to the merger. In this case, it is the following nasal
that conditions the merger of /ɪ/ and /ɛ/. The same vowels in other phonetic
environments (non-nasal) are not necessarily affected, that is, a distinction is still
maintained between words such as “pig” and “peg.” This section will review the
linguistic and social factors that have led to the PIN/PEN merger.
15
2.3.1 Linguistic Factor: Nasalization Unlike vowels in a non-nasal environment, vowels in a pre-nasal context are subject to
complex acoustic interactions caused by having airflow through not only the oral cavity
but also the nasal cavity as the velum lowers in anticipation of the nasal consonant. It
has been found that in American English velar lowering coincides with the onset of the
preceding vowel, meaning that pre-nasal vowels in American English are nasalized
throughout their production (Solé 2007), so the acoustic effects of nasalization are also
realized throughout the vowel. Nasalized vowels are characterized by “a broader
frequency region in the vicinity of the first formant over which the spectral energy is
distributed for a nasal vowel” (Hawkins & Stevens 1985: 1560). Feng and Castelli (1996),
in their review of previous work, say that for the velar nasal consonant, what they
propose as the target of nasal vowels, typical first and second peaks are at 300 Hz and
1000 Hz, and there is also commonly a low resonance frequency around 250 Hz. The
broader frequency region and interference introduced by the nasal resonances around
F1 means that hearers will have a difficult time accurately perceiving the height of the
vowel, the primary distinction between /ɪ/ and /ɛ/.
The added difficulty of articulating a distinction between /ɪ/ and /ɛ/ introduced
by the nasal resonances may lead speakers to suspend the distinction in favor of ease of
articulation and simply merge the two vowels before nasal consonants. Once the
merger has been established on linguistic grounds, the variant is ready to spread out
from the original innovators to the broader community of speakers.
16
2.3.2 Social Factor: Path of Diffusion Because the PIN/PEN merger is associated with Southern speech, it is appropriate to
consider the spread of the merger in the South. Brown (1990, 1991) conducted an
extensive study of the merger in the South with the oldest data coming from the
journals of civil war soldiers in Tennessee. In Tennessee, she found that the raising of
/ɛ/ to /ɪ/ was present but not common in most of the state for most of the 19th century.
However, a widespread urbanization movement in the South around 1875 was
accompanied by an increased incidence of the merger. Thus, it appeared that the
merger was brought to the cities from rural areas, a type of diffusion known as
contrahierarchical. By the 1930s the merger was nearly complete and ubiquitous in the
state.
Given the present even distribution of the merger across the South, as seen in
Figure 2.4, the merger has been quite robust. There is even evidence, as Labov et al.
(2006) point out, that the merger is actually continuing to spread beyond the traditional
boundaries of the South.
Figure 2.4: The PIN/PEN merger in the North America (Labov et al. 2006)
!"#$%&'"%()#%'*+,+-./012/003/044567"86#*+5#%'+,+.9.90:+3;.2+<=
17
Figure 2.4 shows that the PIN/PEN merger is widespread in the South. The solid red line
shows the extent of monophthongal /ay/ production, one of the most common features
of Southern speech. The hashed line shows the extent of the PIN/PEN merger, which
reaches beyond the border of monophthongal /ay/ and includes portions of southern
Indiana, Illinois, Missouri, Kansas, and Oklahoma.
2.4 Oklahoma: Culture and Language This section gives the relevant history of Oklahoma and discusses how the state has
developed culturally and linguistically since settlement.
2.4.1 Settlement History
Oklahoma became the 46th state to enter the union in 1907. Since then, the area of
the state was split between the Oklahoma Territory to the west and Indian Territory to
the east. Most of the state had been allocated to various Native American tribes
beginning around 1830 after the signing of the Indian Removal Act. The earliest and
densest settlement took place in the eastern part of the state, which included the land of
what are called the Five Civilized Tribes: the Cherokee, Choctaw, Chickasaw, Creek,
and Seminole nations (see Figure 2.3).
18
Figure 2.5: Map of Oklahoma and Indian Territories prior to 1889 (Wikipedia).
After the Civil War, several of the Indian nations were displaced from their lands
as a form of reparations for siding with the Confederacy. One portion of these lands in
the central part of the state became known as the Unassigned Lands as they were not
assigned to any of the other Indian nations. No one could legally settle in the
Unassigned Lands, although several groups attempted to.
Eventually, pressure from these groups of hopeful settlers encouraged congress
to pass an amended form of the Indian Appropriations act of 1889 which opened up the
Unassigned Lands for settlement. Over the next few years, all of western Oklahoma
was opened for Anglo settlement. Figure 2.6 shows the years in which the designated
lands were open for settlement. Most of the lands were settled by “land run,” but one
region was settled by lottery.
19
According to the rules of these land runs, the land to be settled was cleared of all
earlier settlers, all of whom were there illegally, then the land would open up for legal
settlement at a pre-established date and time, usually noon. When the gunshots were
fired at noon, potential settlers could run out and claim the first parcel of 160-acres that
hadn’t already been claimed. On April 22nd, 1889, the first of several land runs occurred
in the Unassigned lands, and a large part of central Oklahoma was settled in a single
day.
Table 2.1 shows the states of origin of those who participated in that first land run
in 1889. Roark (1979) summarizes the provenance of these early settlers saying that
“42% of the population of the Territory came from five Midwestern states and that 21%
of the population came from three Upper Southern states. This suggests that a
Midwestern cultural imprint was significant during the formative period” (in Southard,
1993, p. 237).
234 SOUTHARD
A complete analysis of the LAO data may help resolve these confusing pictures, presenting a clear depiction of dialect distribution which may assist cultural geographers in better defining Oklahoma's place within the regional cultures of the United States. Until that analysis is finished, however, preliminary studies of LAO data suggest an intricate correlation between the settlement history of Oklahoma and its current dialect distribution. The LAO, an autonomous regional project which is part of the larger Linguistic Atlas of the United States and Canada, is particularly significant in identifying dialect distribu-tion within Oklahoma. It represents the most complete and detailed study ever conducted of language within the state. Moreover, it differs in three important ways from other American atlas projects: (1) all data were collected within a four-year span (1960-63) by a single inter-viewer, W. R. Van. Riper, using a consistent interview style; (2) all interviews were recorded on tape and the orig!nal tape recordings were
" placed with the Library of Congress so that other scholars are able to corroborate phonetic transcriptions and calibrate their own transcrip-tions with those for the LAO; (3) all interviews were transcribed by a single, highly competent phonetidan, Raven I. McDavid, Jr., whose interviews and transcriptions constitute a substantial body of the data collected for all the Atlas projects and whose participation ensured a continuity with the other Atlas projects.
Constituting a separate project within the framework of the Lin-guistic Atlas of the United States and Canada, Oklahoma is particularly important to dialect study, for it may provide a "laboratory case" for the study of language interaction when a sparsely populated area is rapidly settled by a new linguistic stock representing various dialects of one language. With the land runs which began in the late nineteenth century, the western half of Oklahoma ("Oklahoma Territory") was settled literally overnight by English-speaking peoples. The eastern half of the state, which constituted "Indian Territory" at the time of the land runs, soon had an Anglo admixture joining the Indian tribes which had been resettled predominantly from southeastern states. The Anglo settlers in Indian Territory followed more traditional migratory paths and timetables, gradually moving in from Missouri, and Arkansas and Texas in particular.
Map 15.6 depicts the land divisions within Oklahoma immediately prior to statehood in 1907. Indian Territory encompassed land reserved exclusively for Indian nations, particularly the "Five Civilized Tribes" (the Cherokees, the Chickasaws, the Choctaws, the Creeks, and the Seminoles) whose allotments are identified by tribal name on map 15.6.
Earlier, from 1830 to 1855, virtually all of Oklahoma save the
PANHANDLE (1890)
Midwestern Speech in Oklahoma
Map 15.6. Oklahoma land divisions prior to 1907.
235
. Panhandle was dedicated by treaty to the Five Civilized Tribes. At the time of the Civil War, however, these tribes signed treaties of alliance with the Confederate States of America and even sent more than 5,000 men to fight on the side of the Confederacy (Gibson 1980: 120). Following the war, the tribes were forced to sign Reconstruction treaties with the United States. These treaties provided for the cession of tribal lands to the federal government; the Seminoles ceded virtually their entire domain (2.17 million acres) to the United States for fifteen cents per acre (Gibson 1980: The other tribes were also forced to cede their westernmost lands, which supposedly were to be used to settle tribes from other parts of the United States.
Ultimately, pressure from Anglos who wanted access to cheap land led to these ceded lands, which constitute most of Oklahoma Territory, being made available via land run for settlement. The dates on map 15.6 identify the areas opened by land run from 1889 to 1893. The Panhandle, as mentioned above, was added to Oklahoma Territory by Congress in 1890 and was open at that time to settlement by home-steaders; it was divided into its present three counties at conferral of statehood in 1907. The area identified on map 15.6 as" Annex" was the former Greer County of Texas. The Supreme Court ruled in 1896 that
Figure 2.6: Oklahoma land divisions prior to 1907 (from Southard 1993)
20
Table 2.1: State of origin of settlers of 1889 Land Run (In Southard 1993, citing Roark 1979)
It is unclear which states Roark considers to be Midwestern and Upper Southern
states, or if he takes into account that the southern portions of Illinois and Indiana
might be better defined as Upper Southern than the more culturally Midwestern
inhabitants further north in those states. However, the important point is that
Oklahoma was settled by people from various parts of the Midwest and the South, and
that the cultural balance probably tilts in the direction of the Midwest, at least for the
1889 settlers who went on to found some of the most important cities in the state:
Oklahoma City (state capitol), Norman (University of Oklahoma), and Stillwater
(Oklahoma State University).
2.4.2 Where is Oklahoma?
The regional and linguistic identity of Oklahoma is unclear as it is situated at a
geographic crossroads between the Midlands region, the South, and the West. Any
number of maps produced by cultural or linguistic geographers will quickly indicate
the mixed regional feelings of the state. This Remarkable Continent (Rooney et al. 1982)
depicts a number of maps that illustrate Oklahoma’s ambiguous identity. A map of
“popular regions of North America” based on the number of times regional terms
appear in telephone directories shows that Oklahoma can be included in four different
regions: the Midwest, the South, the Southwest, and the West (see Figure 2.7).
236 SOUTHARD
the Prairie Dog Fork of the Red River, and not its North Fork, constituted the boundary between Texas and Oklahoma Territory and that Greer County be added to Oklahoma Territory; longtime settlers were al-lowed to file homesteads of 160 acres and to purchase additional land for one dollar per acre (Gibson 1980: 181). Finally, "Lottery" identifies the area of the Kiowa, Comanche, and Apache reservations opened to settlement in 1901 by lottery rather than land.ll;,J.lll.
Little is known of the type of English spoken or the extent to which it was spoken by the Five Civilized Tribes in Indian Territory prior to statehood. Full bloods apparently tended to remain isolated, speaking their native languages and leading a subsistence type of life, relying on· hunting, fishing, and small crops for their livelihood. Their way of life gave rise to the derogatory term blanket Indian, which one may still encounter in Oklahoma. The mixed bloods, however, tended to emu-late Southern whites, developing large farms and plantations in fertile river valleys and even owning slaves. Because the Indians' original homelands were in what came to be the Confederate States and because the mixed bloods emulated numerous elements of Southern culture, even sending their children to private schools and colleges in the South, one might reasonably expect their language to be generally Southern. The migrants from Texas, Arkansas, and southern Missouri who entered Oklahoma following statehood undoubtedly added to the development of a Southern or South Midland dialect in the former Indian Territory.
In contrast to the fuzzy picture of English in Indian Territory, a great deal is known about the settlers of Oklahoma Territory. Much of this information is summarized in Michael Roark's 19 79 dissertation, "Okla-homa Territory: Frontier Development, Migration, and Culture Areas;" which contains an extensive study of the origins of migrants to Oklahoma Territory. Although there is some variation from land-run area to land-run area, Roark's analysis of 1900 census data disclosed the following proportions for state of origin of settlers in Oklahoma Territory as a whole: "Lower Midwest 47o/o, Upper Midwest and North-east 5o/o, Upper South 30o/o and Texas-Lower South 17o/o. Of the four states with the largest number of migrants three were adjacent to the Territory: Kansas (19o/o), Missouri (15o/o), Texas (11 o/o), and Illinois (9o/o)" (131).
An examination of the land-run area of 1889 discloses particularly pertinent information about this earliest Oklahoma Territory settle-ment. The 1890 census, taken one year after the 1889land run, shows a total population in the area of 53,822.3 Table 15.1 shows the distribu-tion by state of origin for the nine states contributing the greatest
Midwestern Speech in Oklahoma 237
Table 15.1 State of Origin of Settlers of 1889 Land Run
Kansas 10,048 19% Missouri 7,421 14% Texas 5,381 10% Illinois 5,347 10% Indiana 4,090 8% Ohio 3,734 7% Iowa 3,003 6% Kentucky 2,895 5% Tennessee 2,507 5o/o
number of settlers. Population figures are followed by approximate percentage of total population. Roark points out that "grouping the states together shows that 42o/o of the population of the Territory came from five Midwestern states and that 21 o/o of the population came from three Upper Southern states. This suggests that a Midwestern cultural imprint was significant during the formative period" (1979: 96).4
In all of the land-run areas, moreover, the settlers were generally from the Midwest, as table 15.2 discloses, although there was some variation from land run to land run. The 1893 Land Run, for example, was predominantly composed of settlers from the Lower and Upper Midwest. Settlers in the Land Runs of 1889 and 1891 tended to divide into northern and southern halves. Compare Cleveland County, at the southern extreme of the 1889 Land Run area, with the other counties formed from that Land Run, for example. The 1891 Land Run shows a considerable disparity between the· northern Lincoln county and the southern Pottawatomie. The 1892 Land Run area tends to divide into eastern and western halves, with such sparsely populated western counties as Roger Mills (6, 190) having a considerably greater percent-age of Southerners (79.8o/o) than the more densely populated eastern counties, such as Blain (10,658) with 34.2o/o from the South. The figures in table 15.2 represent percentages of total population and accordingly do not reflect density of population. Logan County, with a 1900 population of 26,563, for example, had a population more than four times that of Roger Mills County. Accordingly, these data give only approximations as to the influence of different areas of origin.
Land settlement in the Panhandle, Greer County, and the Lottery lands followed a different pattern than did settlement for the areas opened by land run. In both the Panhandle and Grter County, settlers had occupied the lands prior to their addition to Oklahoma Territory, and thus followed more traditional settlement patterns. In the lands
21
Figure 2.7: Popular regions of North America based on telephone book data (Rooney et al. 1982)
The map only includes data from three cities in Oklahoma, which only indicate
the Southwest and the Midwest; however, the justification for the South comes from
data collected in Fort Smith, Arkansas, a city that borders Oklahoma to the east. The
West has been extrapolated seemingly by exclusion since the Midwest seems to end
halfway through Oklahoma, and both Amarillo, Texas and Denver, Colorado are
associated with the West. A number of other maps in the collection corroborate the
mixed, if not confused, regional identity of Oklahoma.
2.4.2.1 Language
Linguistic maps of the United States also illustrate the mixed identity of Oklahoma.
The Atlas of North American English (ANAE) (Labov et al. 2006) considers data from
only four Oklahoma respondents, a male and female each from Oklahoma City and
Tulsa, but fits them into the broader map of English in America. In the Atlas,
Oklahoma is ostensibly presented as belonging to the Midland region (Figure 2.5).
22
Figure 2.5: An overall view of North American Dialects (from Labov, Ash, & Boberg, 2006)
The isogloss for The South encompasses the southern part of the state, but as there were
no respondents from this region, this depiction can only be taken as an extrapolation.
The Oklahoma respondents are classified as speakers of the Midland dialect on the
basis of the criteria determined for that region:
• fronting of /o/, the vowel in “boat”;
• /ɑ/ and /ɔ/not clearly distinct or clearly the same in production and perception,
that is, the COT/CAUGHT merger is in a transitional state;
• no monophtongization of /aj/ before obstruents, a feature that clearly
distinguishes the dialect from the South
(adapted from Labov, Ash, & Boberg, 2006, p. 135)
In addition, the ANAE also indicates an incomplete merger of PIN and PEN in Oklahoma.
In the case of this particular merger, Oklahoma tends to follow the South (Labov et. al,
2006: 68).
!"#$%&'"%()#%'*+,+-./012/003/044567"86#*+5#%'+,+.9.90:+00;44+<=
23
The fronting of /o/, according to the ANAE findings, was most pronounced in
the central Midwestern states Ohio and Indiana. The Oklahoma speakers were
relatively less fronted. The COT/CAUGHT merger was also found to be in a transitional
state in Oklahoma, and a preliminary study of this merger in the present data set
further supports a transitional state in Oklahoma. A COT/CAUGHT distinction is
typically maintained in the South, and given Oklahoma’s settlement history, the
distinction most likely arrived with some of the early settlers. /aj/
monophthongization was clearly absent from the ANAE respondents in Oklahoma, but
such a production is easily found in the state, even if it is less likely in urban areas.
Based on the analysis provided in the ANAE, Oklahoma appears to be distinctly
Midlands but with some Southern features present in transitional states. It is important
to note, though, that the ANAE analysis is based on only four speakers from Oklahoma
City and Tulsa, the two largest cities in the state. The linguistic situation in Oklahoma
is certainly more complex than a study with the scope of the ANAE can indicate.
2.4.3 PIN/PEN in Oklahoma
This section provides a review of linguistic work done in Oklahoma with a focus
on the PIN/PEN vowels. Although several studies of English in Oklahoma have been
undertaken, only Bailey’s Survey of Oklahoma Dialects (SOD) in the early 1990s has
produced a phonological study. E. Bagby Atwood’s The regional vocabulary of Texas
(1962) did include a number of respondents from Oklahoma, but the goal of the project
was to study lexical rather than phonetic variation. William Van Riper began the
Linguistic Atlas of Oklahoma (LAO) in the 1960s. Although he conducted at least 50
interviews between 1960 and 1963, and though many, if not all, of the interviews have
been transcribed by Raven I. McDavid, Jr., no acoustic or phonological analysis of this
24
data has been undertaken to date. Finally, the most recent study conducted in
Oklahoma, Dennis Preston’s Research on Dialects of English in Oklahoma (RODEO), is
an ongoing project that is currently being analyzed by Jon Bakos (2013). The present
thesis also considers a portion of the RODEO data from speakers in and around
Oklahoma City.
As no other data has yet been thoroughly analyzed besides SOD, that study is
the basis for this review of Oklahoma. SOD was an extensive study that utilized a
relatively quick but geographically broad telephone survey along with a more in-depth
field survey. The telephone survey gathered responses from 632 people, and the field
survey collected data from 145 respondents (see Bailey et al. 1997). Data was collected
systematically across the state from a full range of age groups in order to allow for an
apparent time study that is representative of the entire population of the state.
As part of the SOD study, respondents were assessed for their production of
“pen” and “Wednesday”. The traditional form was considered [ɛ] and the innovative
pronunciation raised the vowel to [ɪ]. Tillery (1997) found a number of social factors to
be relevant to the raising of /ɛ/ to /ɪ/ in Oklahoma, which are summarized in Table
2.2.
Linguistic Variable
Age Gender Occupation Education Nativity Years in Neighborhood
Rurality Size of longest Residence
/ɪ/ in Wednesday
__ __ __ .01 .01 __ .01 .01
/ɪ/ in pen
__ __ __ .01 .01 .01 .01 .01
Table 2.2: Significant correlations (p) between linguistic features and demographic information in SOD (Tillery 1997)
The factors found to be significant include level of education, whether the
speaker was born in Oklahoma or not, whether the speaker lived in an urban or rural
location, and the size of the place where the speaker had lived the longest. This final
25
factor is related to the rurality factor. It is interesting to note here that age and gender
are not found to be significant for the PIN/PEN merger. This would suggest that as of
the early 1990s the merger is no longer in progress and that it has instead settled into a
marker of social identity.
SOD data indicates that the PIN/PEN merger nearly went to completion in the
1930s (Bailey et al. 1995), just as Brown (1991) found in Tennessee, suggesting that
similar phonological and social processes were at work in both states. However, the
merger apparently began to reverse, especially after WWII. As mentioned in section
2.1.2, mergers can only be reversed by non-linguistic processes. Bailey et al. (1995)
suggest that several factors led to the reversal of the PIN/PEN merger. The
establishment of several military bases and a strong economy brought many new
people to the state, and these immigrants, they claim, were more likely to distinguish
the two vowels. Many of these new immigrants were in the upper-middle class and
settled in the suburbs of Oklahoma City and Tulsa, areas that were influential in the
diffusion of linguistic change.
3 Methodology: Data Collection This study considers the production of vowels in the PIN/PEN classes for 24 speakers of
Oklahoma English living in Oklahoma City and the surrounding rural areas. Wordlist
data was evaluated impressionistically and acoustically to determine if individual
speakers produce a distinction between minimal and near-minimal pairs. An acoustic
analysis also documents the change in speakers’ vowel quality over time.
26
3.1 Participants
For this study, data was collected from 24 speakers. Speakers were given pseudonyms.
There were three speakers in each of eight categories, defined by age, “rurality,” and
gender. These factors were defined as follows:
Age
Respondents were classified as “young” if they were between 18 and 35, and they were
classified as “middle,” or middle-aged, if they were between 45 and 65.
Rurality
Respondents were classified as “rural” if they came from a town of fewer than 25,000
people or from a rural area. All of the rural respondents lived in areas close to
Oklahoma City, and all identified with being “rural” when asked directly. “Urban”
respondents had lived in Oklahoma City their entire lives, with the exception of leaving
for college.
Gender
Respondents were either “male” or “female.”
Several demographic factors were held constant among all respondents,
including nativity, education, and general geographic area. Participants in this study
were all natives of Oklahoma or had moved to the state before the age of 4. Some
respondents had lived outside of the state for a predetermined amount of time, such as
for college or professional training. As for education, all participants held at least an
associate’s degree. Finally, all participants were from the same general geographic
region, being Oklahoma City and the surrounding rural areas.
27
The birthplace and native language of both parents of the participants was noted,
but this information did not affect the selection of participants. None of the
participants spoke another language at home besides English.
3.2 Data collection
Data for this study was collected as part of a longer interview which consisted of
five parts: 1) an elicitation of casual speech, 2) a perceptual dialect map-drawing task, 3)
a reading passage, 4) a word list, and 5) an auditory perception study. This thesis only
considers data from the word list, as it is likely to be the most formal off all of the
speaking styles elicited in this study, as discussed below.
3.2.1 Interview
Interviews were conducted following the general protocol for sociolinguistic
interviews described in Labov (2006). A number of different speech styles were elicited
from participants in order of increasing formality, as Labov outlined. Respondents
were first given a general overview of the nature of the research, which was to study
the way that Oklahomans talk. They were told that there were no right or wrong
answers to any of the questions or tasks, and that the purpose was simply to see how
they really talk. Just before the interview was begun, the consent form was explained to
them and they were asked to sign it. At all times an effort was made on the part of the
researcher to be transparent.
This thesis only considers data from the word list, as it can be considered the most
formal style in the interview according to Labov (2006). In the production of formal
speech, speakers are more likely to pronounce words carefully and in keeping with
their conception of prescribed norms. Concerning the PIN/PEN merger, as we have seen,
phonetic factors are driving the merger (generally a Southern feature), but social factors
28
(e.g. prestige) support the split (a non-Southern feature). Southern features are typically
stigmatized in American speech (see, for example Preston 1996) so anyone who does
not want to sound Southern should try to avoid Southern features, including the
PIN/PEN merger, if they are able to do so. As prescribed norms exclude stigmatized
features almost by definition, and because formal speech styles are more likely to elicit a
speaker’s conception of prescribed norms, the word list reading, being a formal task,
can be regarded as the most likely to provide evidence of a split vowel system, if it
exists. If the vowels are not distinguished in this formal style, then it is likely they are
also not distinguished in less formal registers. We can conclude that if a speaker
merges the vowel pairs here, the speaker is certainly merged. Using word list data, we
can thus be as assured as possible that the evidence we have available points to one
state or the other.
3.2.2 Wordlist
The tokens considered in this study were elicited and recorded as part of a longer
wordlist (approximately 100 words, see Appendix A) designed to study other aspects of
Oklahoma English. Words were presented to participants in the form of a slideshow on
a laptop computer. The words appeared individually and were read aloud one time by
each participant. To minimize the effects of list intonation, the interviewer manually
controlled the rate of presentation so that words were presented at uneven time
intervals. Seven tokens in the PIN/PEN class were elicited, as presented in Table 3.1.
29
PIN/PEN
/ɪ/ /ɛ/
tin ten
sinned send
windy Wendy
dim hem
thing then
Table 3.1: PIN/PEN vowels elicited in wordlist, with minimal pairs in bold
As shown above, this list included several minimal pairs, to allow for direct
comparison of vowels. The top two minimal pairs are the primary focus of this study as
they provide the clearest examples of a merger or split since the vowels in question are
in the same phonetic environment and cannot be influenced in different ways by the
surrounding consonants. The “wind/Wendy” pair is only considered in the
impressionistic evaluation because the adjacent glide makes acoustic analysis difficult.
Even in the impressionistic evaluation, the pair is considered less indicative of a merger
or split due to the influence of the glide. “Then” is not explicitly considered in this
paper because a preliminary analysis found that it tended to follow the pattern of
“hem” and a second near-minimal pair was not available for a systematic comparison.
One token with a velar nasal was elicited, “thing”. This word was ultimately excluded
from analysis as many speakers produced the vowel in “thing” differently than the
other PIN/PEN vowels; in “thing” the vowel was produced more like /e/. This
production is common in Oklahoma, and words such as “thank” and “think” are often
homophonous in Oklahoma speech. Although these vowels before velar nasals thus do
30
appear to be involved in a merger, it is not the merger under consideration in this
study.
The wordlist also included a set of reference vowels in the environment /hVd/,
as shown in Table 3.2. These are the same 12 vowels in the same context as were elicited
and studied in Hillenbrand, Getty, Clark, and Wheeler (1995), which is based on
Peterson and Barney (1952).
Vowel Word /i/ heed /ɪ/ hid /e/ hayed /ɛ/ head /æ/ had /ʌ/ HUD
/ɑ/ hod /ɔ/ hawed /ɚ/ heard /o/ hoed /ʊ/ hood /u/ who’d
Table 3.2: hVd words presented in the wordlist
The vowels produced in this context serve as helpful reference points because they
indicate the formant frequencies of some of the most common vowels in American
English. They can be used to show the extent of a given speaker’s vowel space. Of
particular importance to our examination of the PIN/PEN vowels, these words can also
tell us where speakers produce /ɪ/and /ɛ/ when they are not influenced by a following
nasal consonant.
31
3.3 Recording and Preparation for Analysis
Participants were recorded in the field, usually in their homes or at a local
community center, using a Shure Beta 53 omnidirectional condenser headworn vocal
microphone routed through a Sound Solutions USBpre amplifier. The microphone was
positioned approximately one inch to the side of the respondent’s mouth. Recordings
were made using Praat and were collected at a sampling rate of 44,000 Hz. The
recordings were resampled to 10,000 Hz prior to analysis in order to reduce the file size
and speed up analysis.
An effort was made during recording to maximize the signal-to-noise ratio by
selecting quiet locations and avoiding sources of low frequency background noise, such
as refrigerators and air conditioner vents. As recordings were made directly into Praat,
a laptop computer was present during recording and was placed either on a table or the
floor, depending on the recording environment.
The lightweight, headworn microphone was chosen because it is relatively
inconspicuous and remains at a consistent distance from the respondent’s mouth.
Respondents did not show overt signs of being bothered by the presence of the
microphone.
Recordings were typically made in two back-to-back sessions and lasted from 45
to 75 minutes in total. The first session consisted of the “casual” conversation
component of the interview. In this session, respondents were asked to provide some
basic background information, including date and place of birth, residence history,
provenance of parents, education, perspectives on Oklahoma, etc. This portion also
included a map-drawing task in which respondents were presented with a “blank” map
of Oklahoma and asked to indicate where people speak differently in the state (see
32
Preston, 1981). This “perceptual dialectology” component was intended to elicit
respondents’ attitudes and beliefs about English in Oklahoma, but this information is
not considered in the present analysis.
At the conclusion of the first portion of the interview, the recording was stopped
and saved. The second portion proceeded immediately afterwards unless the
respondent wished to take a short break, which rarely happened. The second portion
consisted of a reading passage and word list. After the word list had been read, the
recording was stopped and saved. A perception experiment intended to determine
respondents’ ability to distinguish several potential vowel mergers (PIN/PEN,
COT/CAUGHT, vowels before /l/) followed but also is not considered here.
4 Data Analysis and Discussion This section presents the results of the impressionistic evaluation and acoustic analysis
of the wordlist data.
This study first categorizes the speakers’ production of the PIN/PEN vowels
according to degree of merger or split based on an impressionistic evaluation (Section
4.1). Speakers are assessed relative to their own production and not based on
conformity to a predetermined “norm.” Next, an acoustic analysis further explores the
various production methods speakers use (Section 4.2). The acoustic analysis looks
primarily at the shape of F2 during the vowel as it is the clearest indicator of variation
between speakers. A summary of the linguistic findings from the impressionistic
evaluation and acoustic analysis is given in 4.3. Finally, the social distribution of
variables identified in the impressionistic and acoustic analyses is considered (Section
4.4).
33
The findings show that Oklahomans not only have the option to merge or not
merge PIN and PEN, but they also have two or three options for how they merge or do
not merge the vowel classes – those with a relatively steady or slightly downward F2
trajectory, those with a humped F2 trajectory indicative of triphthongal production, and
those who mix the two production styles. Section 0 will provide examples of each of
these options.
4.1 Impressionistic Evaluation An impressionistic evaluation of several minimal pairs and a near-minimal pair was
done in order to determine if the speakers produced a recognizable difference in the
production of the (near-) minimal pairs.
4.1.1 Methods Two evaluators with training in phonetics listened to the minimal and near-minimal
pairs (tin/ten, sinned/send, windy/Wendy, and dim/hem) and rated the degree of
similarity or difference between the pairs relative to the speakers’ own production.
Although many token pairs were clearly similar or different, in cases where pairs were
not clearly similar or different, a third opinion was sought.
Token pairs were designated as “same” or “different.” Based on these results
from production of minimal pairs, speakers were classified according to degree of
merger. If all four pairs were determined to be “same” or “different,” they were
considered “strongly merged” or “strongly split”, coded as “M1” and “S1,”
respectively. Speakers were considered “moderately merged” or “moderately split” if
one minimal pair was different from the determinations of the other pairs, coded as
“M2” and S2,” respectively. Speakers for whom the “Wendy”/”windy” pair was the
only pair inconsistent with the determinations of the other minimal pairs were still
34
considered strongly merged or split since the preceding glide often interfered with the
production of the following vowel. The results of the impressionistic evaluation are
given in Table 4.1.
4.1.2 Results
Speaker Age Location Gender SEND/SINNED TEN/TIN WENDY/WINDY HEM/DIM Status
19 Allen M U M same same same same M1 4 Amber Y R F different different same different S1 22 Chad Y U M different different different different S1 7 Cody M R M same same same same M1 8 Dale M R M same same same same M1 5 Danny Y R F same same same same M1 23 Fitz Y U M different different different different S1 10 Gabe Y R M same different same same M2 9 Herbert M R M same same same same M1 13 Hilda M U F different same same same M2 20 Jack M U M different different different different S1 11 James Y R M different different different different S1 14 Jane M U F same same different same M1 24 Mason Y U M different same same different M1* 1 Pam M R F same same different same M1 15 Janet M U F same same different same M1 16 Ramona Y U F different different different different S1 2 Renee M R F same different different different S2 17 Shelly Y U F different different different different S1 21 Sam M U M different different same same S2 18 Skylar Y U F different different different different S1 6 Sarah Y R F same same same same M1 3 Tracy M R F different different different different S1 12 Van Gough Y R M same different same same M2
Table 4.1: Results of Impressionistic Evaluation (M1 = Strongly merged, M2 = Moderately merged, S1 = Strongly split, S2 = Moderately split)
Many of the speakers, eight out of twenty-four, consistently produced very clear
distinctions while others, six out of twenty-four, consistently produced the minimal
pairs in very similar manners. Of those for whom the “Wendy”/”windy” pair was the
only one differing in evaluation from the other three, one (Amber) merged the pair
35
when she had an otherwise split system, and three (Jane, Pam, and Janet) actually
distinguished the pair even though they had an otherwise merged system. The only
other token pair that yielded evaluations that diverged from the speaker’s other three
pairs was “send”/”sinned,” including two speakers, Hilda – who produced a
distinction in an otherwise merged system, and Renee – who produced the two tokens
the same in an otherwise split system.
Finally, two speakers produced half of the pairs similarly and the other half
differently, Mason and Sam. These two speakers’ production, however, is less easy to
describe according to the guidelines used for the other speakers. Sam produced weak
distinctions even for the pairs that were evaluated as different. For Mason, although
individual token pairs may have been either “same” or “different,” his means of
producing the vowels across all four pairs was not consistent. He produces “dim” with
[ɛ] and “hem” with [ɪ] and similarly inverts the vowels in “sinned” and “send.” “Tin”
and “ten” and “Wendy” and “windy” are produced the same, with [ɪ]. Without more
data it is difficult to offer an explanation for Mason’s production; however, this sort of
mixed production that blurs the line between /ɪ/ and /ɛ/ is what would be expected in
a merger-by-expansion (see Herold, 1990), that is to say, a phonological merger with the
allophones [ɪ] and [ɛ] in free variation. It is also interesting to note that Mason is the son
of Jane, a strongly merged speaker, and Sam, the moderately split speaker. Exposure to
these differing vowel systems, in light of Herzog’s Principle that mergers expand at the
expense of distinctions, provides a simple explanation for Mason’s merged state.
Speculating further to suggest an explanation for Mason’s inverted production of the
vowels in question, he may have learned the two distinct vowel targets from his father
and others with a similar system, and inherited an enlarged phonemic space from his
36
mother and other merged speakers. If this were the case, Mason’s may be a transitional
state on the way to a complete merger-by-transfer or merger-by-approximation.
4.2 Acoustic Analysis This acoustic analysis focuses primarily on a visual inspection of the spectrograms.
Although measurements will be given to the extent that such measurements can be
considered valid will the given when available, the nature of the particular class of
vowels in question makes the taking of accurate and valid measurements difficult.
First, the problem of nasalized vowels is discussed, followed by a presentation of the
various sorts of formant contours found in the data set. Then, examples of intraspeaker
variation are given to show how several types of formant contours can be present in a
single speaker’s production.
4.2.1 Nasalized Vowels Unlike vowels in a non-nasal environment, vowels in a pre-nasal context are subject to
complex acoustic interactions caused by having airflow through not only the oral cavity
but also the nasal cavity once the velum lowers in anticipation of the nasal consonant.
Nasalized vowels are characterized by “a broader frequency region in the vicinity of the
first formant over which the spectral energy is distributed for a nasal vowel” (Hawkins
& Stevens 1985: 1560). Feng and Castelli (1996), in their review of previous work, say
that for the velar nasal consonant, what they propose as the target of nasal vowels,
typical first and second peaks are at 300 Hz and 1000 Hz, and there is also commonly a
low resonance frequency around 250 Hz. Nasalization is also accompanied by an
antiformant in the vicinity of F1, which varies in position depending on the degree of
nasalization. These additional and conflicting, sometimes canceling, resonances pose a
problem because it makes it difficult to be confident about the location of the oral F1.
37
For this reason, this portion of the study will not place a strong emphasis on formant
measurements as many other vowel studies have done in recent years. A method of
data collection designed to capture nasal and oral resonances separately would
certainly reveal interesting details about the role of F1 in these pre-nasal vowels.
4.2.2 Types of formant contours Three types of formant contours are present in the data set analyzed in this study. The
relevant contours are found in F2 of the vowels. F1 evidently plays a role in the
production of these vowels, but interference from the nasal resonances prohibits
consistent analysis of the available data. The first type is referred to as “flat” because the
vowel produces a flat, steady state F2 through the duration of the vowel. This sort of
production results in a monophthong. The second type is referred to as “humped”
because, in the most prototypical of cases, begins fairly high, increases by about 100 Hz
over approximately one-third of the duration of the vowel, then descends to a point
below the onset and may or may not achieve a steady state. This sort of production
results in a triphthong. The third type can be described most generally as a
“downward” slope of F2; however, the steepness and shape of F2 may vary. This sort
of production results in a diphthong. Two kinds of diphthongs can be observed in this
data set, those for which a steady state occurs in the first part of the vowel followed by a
downward slope that may or may not reach another steady state, and those which
consist of only the downward slope. Bigham (2005) identified similar diphthongs in his
study of Southern Illinois, so his terminology will be adopted here, referring to the
former diphthongs as “true diphthongs” and the latter as “breaking diphthongs.”
38
4.2.3 Monophthongs: “flat” countours A prototypical monophthong aims to hit only one target vowel quality that is
maintained for the entire duration of the vowel. Some variation is expected, of course,
and the following three examples present some of the observed varieties.
With an F2 onset at 1735 Hz and an offset at 1700, Fitz’s production of “send” in
Figure 4.1 presents a clear example of a monophthongal production. F1 is also level for
the duration of the vowel.
Figure 4.1: Fitz "send"
Time (s)0.4629 0.9629
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“send”
39
Shelly’s production of “sinned” (Figure 4.2) has an F2 onset at 2150 Hz and an
offset at 2180 Hz but the formant trajectory reaches a peak around the middle of the
vowel at 2280 Hz, 100 Hz higher than the offset. Fluctuations such as these need not be
evidence of a diphthongal pronunciation since there is clearly only one vowel target.
This example also shows how the nasal resonances interfere with the oral resonances in
the area of F1.
Figure 4.2: Shelly "sinned"
Time (s)0.3403 0.8403
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“sinned”
40
Figure 4.3 shows Amber’s gently sloping F2 in her production of “tin.” The onset
is at 2520 Hz and the offset is at 2230 Hz, a difference of about 300 Hz. Formant tracks
have been added to this figure to highlight the oral formant. Productions with such a
change from onset to offset are not as prototypically monophthongal as Fitz’ “send” in
Figure 4.1, but there are certainly more similarities between these two than with any of
the diphthongal examples, especially since Amber’s F1 remains steady throughout the
vowel.
Figure 4.3: Amber "tin"
Time (s)0.4514 0.8374
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
41
4.2.4 Triphthongs: “humped” contours This section presents examples of triphthongs, which pass through three targets within
the span of the vowel. Feagin (1987) presents some examples of triphthongal
production in the South, such as “crib,” transcribed as [khrɪiyəb]. This kind of
production is evidently similar to what can be observed in Dale’s production. Dale
produces “send” (Figure 4.4) with an onset at 1970 Hz, a peak at 2500 Hz, and an offset
at 1680 Hz. The F2 values at the onset and peak correspond with the values of [ɪ] and a
fronted version of [i] as presented in Peterson & Barney (1952) followed by an offglide,
which is often transcribed as [ə]. Because the most prominent target in this type of
production is [i], a tense vowel, this kind of vowel will be referred to as a “tense
triphthong.”
Figure 4.4: Dale "send"
Time (s)0.3436 0.8984
0
5000
Freq
uenc
y (H
z)
0
1
2
3
4
5
42
In Figure 4.5 we see that Allen also produces “send” in a “humped” manner with
an onset at 1680 Hz, a peak at 1900 Hz, and an offset at 1420 Hz. This F2 peak at 1900
Hz is in the vicinity of Peterson & Barney’s (1952) [ɪ]. Because the most prominent
target in this type of production is [ɪ], a lax vowel, this kind of vowel will be referred to
as a “lax triphthong.”
Figure 4.5: Allen "send"
Time (s)0.6693 1.332
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
43
4.2.5 Diphthongs: “sloped” contours Two kinds of diphthongs were identified in this data set, “true” diphthongs and
“breaking” diphthongs. This section will present examples of both.
Sam’s production of “dim” (Figure 4.6) presents a relatively flat trajectory during
the first 40% of the vowel before pursuing a downward trajectory. The onset is at 1750,
there is a slight 55 Hz rise up to the breaking point, and the offset is at 1080. The final
offglide is similar in appearance to the triphthongal offglide in that it continues a
downward trajectory without reaching a steady state before transition into the nasal.
Figure 4.6: Sam "dim"
Figure 4.7 shows Hilda’s production of “hem,” which shows an F2 with steep and
steady downward trajectory with nothing identifiable as a steady state, a breaking
diphthong. The onset is at 2690 Hz and the offset is at 1460 Hz, a difference of more
Time (s)0.6664 1.001
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
44
than 1200 Hz. Also of note is the apparently raising F1 in the latter half of the vowel, a
feature that is consistent with the production of a diphthong.
Figure 4.7: Hilda "hem"
As we will see in the following section, breaking diphthongs occur almost
exclusively in the token “hem,” even for speakers who have monphthongal productions
in their other tokens, suggesting that this sort of production is the result of
coarticulatory effects rather than the phonology of the vowels. There is evidence in the
aspirated portion of tokens with breaking diphthongs (sometimes “tin” and “ten” in
addition to “hem”) that true diphthongs, or even triphthongs, are actually articulated
by some speakers before the onset of voicing.
Time (s)0.4479 0.8903
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
45
4.2.6 Examples of intraspeaker variation Although it is convenient to have such clear examples of each of the varieties of formant
contours identified in the data, not all of the spectrograms present such relatively clean
and unambiguous depictions of the acoustic signal as shown in the pervious section.
Speakers vary in ways that may be attributed to phonetic context, but they also vary
because the production of human speech is variable by nature. This section considers
the speech of one speaker from each of the eight demographic groups surveyed in this
study. A collection of six spectrograms are presented for each speaker consisting of two
minimal pairs and one near-minimal pair: “sinned,” “send,” “tin,” “ten,” “dim,” and
“hem.” These speakers are not necessarily thought to be “representative” of their
respective demographic groups, but it is convenient to show the diversity of variation
of the entire data set by selecting one speaker from each group. The demographic
description of each speaker is given after his or her pseudonym in the form (age,
location, gender) such that “YUM” stands for a young, urban, male, and “MRF” stands
for a middle-aged, rural, female. The speakers’ merger status is also given, as
determined in the impressionistic analysis.
Spectrograms in this section are shown without formant tracks in order to
provide a clear view of each image. Images were created be selecting a frame from 100
ms before and 400 ms after the onset of voicing to give a consistent total frame width of
500 ms.
46
4.2.6.1 Allen (MUM) – M1
Figure 4.8: Allen
At first glance, Allen appears to produce these vowels as true diphthongs with some
coarticulatory effects creating a lax triphthong in his production of “sinned” and
“send.” However, closer inspection of the aspirated portions of “tin” and “ten” reveal
formants in the region of F2 that directly track onto the same formant in the voiced
Time (s)0.5637 1.064
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“tin”
Time (s)0.6066 1.107
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“ten”
Time (s)0.9395 1.44
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“sinned”
Time (s)0.7809 1.281
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“send”
Time (s)0.8742 1.374
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“dim”
Time (s)0.6639 1.164
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“hem”
47
portion resulting in a contour the mirrors that of F2 in “sinned,” “send,” and, to a lesser
extent, “dim.” The voiced portion of “hem” appears to be a breaking diphthong;
however, the aspirated portion has an F2 contour that suggests at least a true
diphthong, if not a tripthongal articulation. There is no substantial difference in length
between the two vowel classes and the minimal pairs produced nearly identical
spectrogram images, corroborating the impressionistic evaluation as strongly merged.
48
4.2.6.2 Fitz (YUM) – S1
Figure 4.9: Fitz
Fitz is clearly and prototypically monophthongal in the first four tokens. The final two
show slightly greater downward trajectories, but the change from onset to offset is
relatively slight and consistent with other monophthongal speaker’s production of
“dim” and “hem.” F1 is higher for all of the PEN-class vowels, supporting the
impressionistic evaluation of strongly split. The fact that oral F1 is visible in the
Time (s)0.3214 0.8214
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“tin”
Time (s)0.273 0.7730
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“ten”
Time (s)0.3749 0.8749
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“sinned”
Time (s)0.4629 0.9629
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“send”
Time (s)0.3071 0.8071
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“dim”
Time (s)0.3688 0.8688
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“hem”
49
spectrogram for PEN-class above the lower frequency nasal resonances suggests that
such a production is important in producing the vowel class distinction. Neither vowel
class appears to be longer than the other.
50
4.2.6.3 Hilda (MUF) – M2
Figure 4.10: Hilda
Hilda shows much variation in her production style, but all closer to the triphthongal
end of the spectrum than the monophthongal end. “Sinned” is produced as a lax
triphthong while its counterpart “send” is a true diphthong. This is a meaningful
difference as this pair was evaluated as sounding different. “Tin” and “ten” are both
diphthongal in the voiced portion, but there is some vague evidence in the aspirated
Time (s)0.4931 0.9931
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“tin”
Time (s)0.5933 1.093
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“ten”
Time (s)0.4342 0.9342
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“sinned”
Time (s)0.4398 0.9398
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“send”
Time (s)0.3772 0.8772
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“dim”
Time (s)0.4865 0.9865
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“hem”
51
portion of a triphthongal articulation. “Dim” appears as a true diphthong while “hem”
has a sharply breaking diphthong that is somewhat characteristic of that token. Hilda is
not strongly merged, and she does, in fact, produce a distinction between “sinned” and
“send,” but the weight of the evidence suggests a moderate merger. Furthermore,
neither vowel class was longer than the other.
52
4.2.6.4 Ramona (YUF) – S1
Figure 4.11: Ramona
Ramona, who also happens to be Fitz’ sister, shares her brother’s means of production.
F2 for three of the first four tokens is quite steady. Even the relatively slight downward
slope in “tin” is minor enough to be considered monophthongal. “Dim” and “hem”
show the slightly downward sloping F2 characteristic of these tokens as produced by
Time (s)0.394 0.8940
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“tin”
Time (s)0.4961 0.9961
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“ten”
Time (s)0.3504 0.8504
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“sinned”
Time (s)0.3932 0.8932
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“send”
Time (s)0.3792 0.8792
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“dim”
Time (s)0.3853 0.8853
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“hem”
53
monophthongal speakers. F2 is higher for the PIN-class vowels in all cases, and the PEN-
class vowels tend to be longer, thus supporting a strong merger evaluation.
54
4.2.6.5 Herbert (MRM) – M1
Figure 4.12: Herbert
Herbert produces lax triphthongs in “sinned” and “send.” As is typical of speakers of
this type, “tin” and “ten” appear to be true diphthongs, but the aspirated portion of the
words suggest them to be articulatorily triphthongal. “Dim” has a contour similar to
that produced in the same token by Hilda (Figure 4.10). “Hem,” in contrast to most of
Time (s)0.3097 0.8097
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“tin”
Time (s)0.2377 0.7377
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“ten”
Time (s)0.3524 0.8524
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“sinned”
Time (s)0.4832 0.9832
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“send”
Time (s)0.2959 0.7959
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“dim”
Time (s)0.4093 0.9093
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“hem”
55
the other speakers’ production of the same token, is a true diphthong rather than a
breaking diphthong. This assessment bears out in the aspirated portion as well since F2
remains flat. The spectrograms of the minimal pairs look nearly identical, supporting
the impressionistic evaluation as strongly merged. Vowel length was not substantially
different between the two classes.
56
4.2.6.6 Van (YRM) – M2
Figure 4.13: Van
As Hilda showed a great deal of variation on the triphthongal end of the continuum, so
does Van show on the monophthongal end of the continuum. Three of the first four
vowels in the figure show downward sloping F2s that toe the line between what can be
considered monophthongs and breaking diphthongs. The apparent raising of F1 towad
the end of the nucleus for the PIN-class vowels also detracts from a monophthonal
Time (s)19.21 19.710
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“tin”
Time (s)0.4679 0.9679
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“ten”
Time (s)0.5268 1.027
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“sinned”
Time (s)0.5892 1.089
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“send”
Time (s)0.6182 1.118
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“dim”
Time (s)0.8323 1.332
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“hem”
57
classification. “Dim” presents a true diphthong while “hem” shows the same
downward trajectory as most other monophthongal speakers. These two tokens have
similar F2 contours when including the aspirated portion of “hem.” Although “sinned”
and “send” were evaluated as different, and “ten” is longer than “tin” Van seems to be
more merged than not. More data could contradict this assessment, but the available
evidence supports this conclusion.
58
4.2.6.7 Renee (MRF) – S2
Figure 4.14: Renee
Renee appears to be a typical tense triphthongal speaker; however, unlike the other
triphthongal speakers, she differentiates the PIN- and PEN-class vowels. “Sinned” and
“send” were evaluated as being the same, but all of the other pairs were different. The
spectrograms reveal that her PIN vowels are produced as tense triphthongs and the PEN
vowels are produced as diphthongs. She is the only speaker in this data set who uses a
Time (s)0.312 0.8120
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“tin”
Time (s)0.3661 0.8661
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“ten”
Time (s)0.4195 0.9195
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“sinned”
Time (s)0.3987 0.8987
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“send”
Time (s)0.1932 0.6932
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“dim”
Time (s)0.3333 0.8333
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“hem”
59
triphthongal production style who is not merged, and she accomplishes this split by
tending to produce diphthongal PEN-class vowels.
60
4.2.6.8 Amber (YRF) – S1
Figure 4.15: Amber
Amber produces the PIN and PEN class vowels differently. The PIN vowels have a higher
F2, shorter length, and a more diphthongal quality than the PEN class vowels. “Dim”
and “hem,” while different in appearance from Amber’s other tokens in the figure,
more closely resemble lax triphthongal speakers’ production of the same tokens than
they resemble monophthongal speakers’ production (e.g. Hilda and Renee). Although
Time (s)0.4804 0.9804
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“tin”
Time (s)0.456 0.9560
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“ten”
Time (s)0.6293 1.129
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“sinned”
Time (s)0.5451 1.045
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“send”
Time (s)0.6168 1.117
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“dim”
Time (s)0.4958 0.9958
0
5000
Freq
uenc
y (H
z)
0
1000
2000
3000
4000
5000
“hem”
61
this section of the results has avoided making comments on demographics, it is
interesting to note here that she is young and rural, and has aspects of both
monophthongal and triphthongal production, suggesting that she is in a transition state
going toward a more urban speaking style. Much more data would be needed to
support this idea of a transitional speaker, but speakers like Amber would make
interesting case studies to see how much and what kind of variant production is present
in transitional speakers.
62
4.3 Summary of Linguistic Findings It is evident that the merger of PIN and PEN in Oklahoma is not a simple matter of
speakers either producing a distinction between the vowel classes or not. There are two
systems interacting. One system includes somewhat monophthongal representations of
these vowels, and the speakers may be either merged or not. The monophthongal
speakers who are merged tend to converge on /ɪ/, but one speaker, Van, does appear to
converge on /ɛ/, and another speaker, Mason, has the two phonemes in apparently free
variation. The other system includes vowels of the PIN and PEN classes that are
triphthongal in varying degrees. Some of these speakers aim for a target in the vicinity
of Peterson & Barney’s (1952) /i/, a tense vowel, while others aim for a target in the
vicinity of /ɪ/, a lax vowel.
Table 4.2 presents a summary of the individual speaker’s production styles based
on the impressionistic evaluation and spectrogram analysis. The code in the far right
column summarizes the production styles.
63
Speaker Age Location Gender Merger status Tense/Lax F2 Shape Summary Code1 Allen M U M M1 lax Triphthong M1T(L) Amber Y R F S1 Monophthong S1M Chad Y U M S1 Monophthong S1M Cody M R M M1 tense Triphthong M1T(T) Dale M R M M1 tense Triphthong M1T(T) Danny Y R F M1 Monophthong M1M Fitz Y U M S1 Monophthong S1M Gabe Y R M M2 Monophthong M2M Herbert M R M M1 lax Triphthong M1T(L) Hilda M U F M2 lax Triphthong M2T(L) Jack M U M S1 Monophthong S1M James Y R M S1 Monophthong S1M Jane M U F M1 lax Mixed M1M/T(L) Mason Y U M M1 Monophthong M1M Pam M R F M1 tense Triphthong M1T(T) Janet M U F M1 lax Triphthong M1T(L) Ramona Y U F S1 Monophthong S1M Renee M R F S2 tense Mixed S2M/T(T) Shelly Y U F S1 Monophthong S1M Sam M U M S2 lax Triphthong S2T(L) Skylar Y U F S1 Monophthong S1M Sarah Y R F M1 Monophthong M1M Tracy M R F S1 lax Triphthong S1T(L) Van Gough Y R M M2 Monophthong M2M Table 4.2: Summary of individual speakers' production styles
Thirteen of the speakers in this data set of 24 produce PIN/PEN vowels in a way
that can be considered monophthongal, nine produce the vowels in a way that can be
considered triphthongal, and the remaining two use both monophthongal and
triphthongal productions.
All of the monophthongal speakers produce lax varieties of the vowels. If the
monophthongal speakers do produce certain tokens with diphthongal qualities, it is
typically the result of a weak offglide toward [ə], a trend that appears to be phonetically
conditioned, especially in the case of “hem.” Eight of the monophthongal speakers
1 The code summarizes the analysis results for each speaker. First, the speaker’s merger status is given using the alphanumeric combination described in the impressionistic evaluation section, then speaker’s F2 shape is given as either “T” for triphthongal or “M” for monophthongal, and finally, if the speaker is triphthongal, a letter appears in parentheses to give the tenseness of the vowel, “L” for lax and “T” for tense. Some speakers had mixed productions, in which case both letters are included in the code.
64
produce strongly split (unmerged) PIN and PEN vowel classes, two are moderately
merged, and three are strongly merged.
Of the 11 triphthongal speakers, including those who demonstrate triphthongal
production in some tokens, seven produce lax varieties and the remaining four produce
tense varieties. All three of the exclusively tense triphthongal speakers are strongly
merged, and Renee is the only tense triphthongal speaker who splits the two vowel
classes. Of the lax triphthongal speakers, four are strongly merged, one is moderately
merged, one is moderately split, and one is strongly split.
The evidence available in this relatively limited data set suggests no less than
five different production possibilities of the PIN and PEN vowel classes: merged tense
triphthong, merged lax triphthong, merged monophthong, split tense triphthong with
monophthong, and split monophthong. A split lax triphthong is also suggested by the
data, but only for one speaker and the split was only determined to be moderate, which
is not strong enough evidence in a data set of this size to be convincing.
Earlier acoustic studies have found or assumed F1 (or vowel height) to be a robust
indicator of a merger or split of the PIN/PEN vowels in Oklahoma and similar dialect
regions (Bailey, Wikle, Tillery, & Sand, 1993; Bigham, 2005). In fact, much of the data
available in the current study agree with these earlier findings; however, the
triphthongal and diphthongal vowel quality of some of the speakers in this study lead
us to question if F2 may be used to distinguish the vowel classes for those speakers,
especially given the acoustic difficulties presented by nasal resonances in these pre-
nasal vowels. The diversity of production styles used by Oklahomans means that a
significant amount of detail is lost of one only considers F1 and ignores F2. The
findings of this present study show that F2 plays an important role in the way that some
speakers distinguish the PIN and PEN vowel classes.
65
4.4 Social Distribution The social distribution of the various linguistic variables identified in the earlier sections
is described here. This data set suggests that age and location of residence (urban vs.
rural) are important factors associated with production style of the PIN and PEN vowels
in and around Oklahoma City. Age, alone, serves an almost absolute role in
distinguishing monophthongal and triphthongal production. Location of residence and
tense or lax triphthong production are associated factors among middle-aged speakers.
Location of residence and merger status are also related factors among young speakers.
Data presented in this section are taken from the linguistic variables summarized
in Table 4.2. Because two speakers were determined to have both monophthongal and
triphthongal production styles, those speakers will be included in counts of both styles
when relevant, so the total number of speakers may appear to be greater than the 24
actually considered in this study.
4.4.1 Single Demographic Variables and Production Style The interaction of single demographic variables with production style is considered first
in Figure 4.16, Figure 4.17, and Figure 4.18. In these figures, the blue columns represent
the diametrically opposed variables of merged and split production while the solid red
columns represent the diametrically opposed variables of monophthongal and
triphthongal production. The hatched red columns represent the component makeup
of the triphthongal column by presenting the number of speakers with lax and tense
triphthongal vowels.
66
Figure 4.16: Comparison of age and production style
In Figure 4.16 we see that all of the young speakers have a monophthongal
production style. Only one middle-aged speaker is exclusively monophthongal and the
other two produced some vowels as monophthongs, but not all. All but one of the
middle-aged speakers use triphthongal PIN/PEN vowels. Among these, four use a tense
variety and seven use a lax variety. From the broad perspective of this figure, the
distribution of merged vs. split speakers is not as clear as the distribution of
monophthongs and triphthongs, but the young age group has more split speakers than
the middle age group.
0
2
4
6
8
10
12
# of speakers
Age and Production Style
67
Figure 4.17: Comparision of location of residence and production style
As seen in Figure 4.17, the number of rural merged speakers is twice as many as
rural split speakers and the urban speaker are slightly more likely to be split than
merged. Urban speakers who are triphthongal use the lax variety exclusively while the
rural speakers are more likely to use the tense variety.
Figure 4.18: Comparision of gender and production style
0 1 2 3 4 5 6 7 8 9
# of speakers
Location and Production Style
0 1 2 3 4 5 6 7 8 9
# of speakers
Gender and Production Styles
68
Gender has no identifiable interaction with production styles as presented in
Figure 4.19. All of the columns on the male side are within one speaker difference in the
related column on the female side. Gender may or may not play an important role in
and around Oklahoma City, but the effects are not strong enough to be apparent in a
sample of this size.
4.4.2 Merged vs. Split and Tense vs. Lax Across Demographic Groups This next section takes a closer look at how linguistic features are distributed across
specific demographic groups in this study.
Figure 4.19: Comparison of demographic groups and merger status
In Figure 4.19 we can see that urban speakers are more likely to be split and that
rural speakers are more likely to be merged, but this association appears to be only
weak. Young urban speakers are more likely to be split than their rural counterparts,
but again, only slightly. This figure shows that having a split or merged PIN/PEN
system isn’t simply a matter of demographic group.
0
1
2
3
YUF YUM MUM MRF YRM YRF MUF MRM
# of speakers
Demographic Groups
Split
Merged
69
The following two figures show that demographic groups help to describe the
distribution of linguistic features only when manner of production is taken into
account.
Figure 4.20: Comparison of monophthongal speakers' merger status
Figure 4.20 presents the merger status of all monophthongal speakers. Only one
middle-aged speaker is monphthongal, while the rest are young. Urban speakers tend
to be split while rural speakers tend to be merged, as was noted in Figure 4.19.
0
1
2
3
MUM YRF YRM YUM YUF
# of speakers
Monophthongal Speakers
Split
Merged
70
Figure 4.21: Comparison of Triphthongal Speakers' usage of tense and lax vowel targets
Because only three of the eleven triphthongal speakers are split, a comparison of
two features that show more interesting distribution is more helpful. As mentioned
earlier, only middle-aged speakers were found to produce triphthongs in this data set.
Figure 4.21 shows that rural speakers are more likely to use a tense variety of triphthong
than a lax variety. In contrast, only one urban speaker uses a tense variety, and the rest
use the lax variety.
While the data available in this sample indicates a strong association between
manner of production and age, there is also a weaker effect of location of residence. If a
difference between male and female production exists, it was not evident in this data
set.
5 Conclusion When this research project was first conceived, the original intention was to study the
PIN/PEN merger in a region that is one of the greatest foci of linguistic innovation in a
state that was settled rapidly by speakers from at least two distinct dialect regions. The
0
1
2
3
MRM MRF MUM MUF
# of speakers
Triphthongal Speakers
Lax
Tense
71
only previously published study of Oklahoma that mentions the same merger was done
by an impressionistic evaluation of two tokens, “Wednesday” and “pen” (SOD; Bailey,
Tillery, & Wikle, 1997). That study, with over 800 respondents and nearly 55 target
features, was very broad in scope, so it was hoped that conducting a narrower,
qualitative study of a single linguistic feature in an influential region would reveal a
more detailed picture of one aspect of the linguistic state of Oklahoma.
The results uncovered in this study and presented in this paper show that there
is much more to the PIN/PEN merger than simply asking if a speaker produces the two
classes in a similar way. Rather, it is also about change in vowel quality over time. A
speaker can produce the vowels monophthongally and be either merged or split, and a
speaker can produce the vowels triphthongally with a tense or lax target vowel.
Triphthongal speakers are almost always merged; one of the split speakers uses a
triphthong for one vowel class, and a monophthong for the other, while the other split
triphthongal speaker only produces a weak distinction.
The young speakers in this study strongly favor a monophthongal production of
the PIN/PEN vowels regardless of their merger status or location of residence. The
middle-aged speakers, in contrast, overwhelmingly favor a triphthongal production.
Such a diametric opposition along age lines suggests a robust change in progress.
Considering, though, that the middle-aged speakers also have two ways of producing
the triphthongs, with a lax target more likely for urban speakers and a tense target more
likely for rural speakers, the young speakers are not simply adopting a different
production style than the older generation, they are consolidating what was earlier two
means of production into one, suggesting a leveling of social distinction and weakening
of the urban/rural distinction.
72
One of the greatest difficulties that face this finding of an apparently monolithic
linguistic shift between younger and older speakers comes from the small sample size.
Coordinating evidence across several demographic groups does help to strengthen
otherwise weak evidence; however, it is possible that an education divide between the
young and middle-aged speakers, slight though it may be, may have significant
implications. As all of the young rural speakers had obtained or were pursuing a
bachelor’s degree, most of the middle-aged rural speakers had associate’s degrees from
institutions close to their hometowns. One middle-aged rural speaker had earned a
doctorate, but she produced lax triphthongs, a pattern that is associated with the
middle-aged urban speakers in this study, not with young people. Nonetheless, the
type of degree is not as important as where one must go to obtain one. Fewer
institutions offer bachelor’s degrees than associate’s, so someone would typically need
to travel further to attend a bachelor’s granting institution. It could be that an increased
importance placed on pursuing a bachelor’s degree in the past 30 years has encouraged
rural young people to consider moving away from their hometowns to attend a four-
year institution. As in Labov’s (1963) study of Martha’s Vineyard, these young people
may be experiencing a wholesale positive orientation toward a dialect they perceive as
being more useful, namely the urban dialect. If a large proportion of rural youth
actually did fit this description, it would be very easy for a six-person sample to
inadvertently exclude rural youth who have stronger positive orientations toward rural
life and speech patterns. However, if the young rural speakers were trying to emulate
urban speech, we would expect to see a lag of one generation. So the rural youth ought
to be speaking like the urban middle-aged speakers, and this is not the case.
Based on the evidence available in this data set, we see that an urban/rural
distinction exists among the middle-aged speakers, but the young speakers are
73
essentially the same. There is a sense in which the middle-aged rural respondents grew
up in areas that were rural during their childhood but which are now suburban, or even
urban, as Oklahoma City has expanded. So two generations could have been raised in
the same town, but that town may have been considered rural in the 1950s and
suburban by the 1990s. In this case, it isn’t the urban/rural distinction that has been
weakened, rather rural has become urban and there are no longer two classes to
distinguish. We can speculate that the young rural and urban respondents speak in the
same manner because they are not actually different groups. More work will need to be
done to assess the validity of these claims, but stark contrasts in production between
middle-aged urban and rural speakers, and between young and middle aged speakers
suggests that further exploration would be fruitful.
The settlement history of Oklahoma with people immigrating from both
Midwestern and Southern states and the continuing mixed regional identity of the state
has certainly contributed to the present day linguistic situation in the state. Although
this study has only been able to provide a glimpse into a narrow section of language in
Oklahoma, it has contributed some insights into how Oklahomans vary in production
of key speech sounds. These kinds of variation are the seeds of change. Understanding
what varieties are available to hearers and speakers and how those varieties are
distributed in the population are part of the necessary background work needed to
inform future studies language variation and change. Many more questions can be
asked, and it is my hope that this study has provided a platform from which to further
explore language in Oklahoma.
74
Appendix A: Wordlist
Tree Pig Wendy Hayed Day Every Jab Tin HUD Cob Saw Hoe Good Who’d Hawk Chew Duty How Don Hoed Boy Lie Those Ruth Wash Business Garage Soda Head Shrimp Strike Houston Floyd Seven With Hock Hod Cloud Sinned Steve Trade Sand
Thing Measure Shop Hug Heat Mesh Thick Strength Peel Talker Tuesday Loan Cut Heed Send Shoot Knife Hook Forty Push Hawed Out Brother Lied Chewed Then Heard Windy Happy Sang Hid Ten Bet Pawed Fail Dim Ate Cool Where Dawn Hood Boat
Mat Hem Fish Wasn’t Had
75
Appendix B: IRB Consent form
Consent Form - How Oklahomans TalkResearchers: Dennis R. Preston and student fieldworkers
This is a study of how people in Oklahoma talk. In this study, we are not at all interested in "right"and "wrong" and would never refer to the results of this research that way. We are fascinated by the waylanguage changes from place to place, and our study will contribute to tl-re scientific knowledge oflanguage and our ability to advise people in education, the law, and other public domains that areconcerned with language. I wor-rld be happy to discuss these objectives with you before your decision toparticipate or during or after the interview.
I will record your speech (and video record the interview if you agree) and ask questions about whatyou think about language. I will ask you to read a list of words and a short passage and ask you aboutyour life history. I will ask you to identify words from a recording and imitate how others speak, if youare comfoftable doing so. Tliis interview should take no more than an hour. Your participation iscompletely voluntary, and you may choose not to participate at all, or you may refuse to participate incertain parts of the interview, refuse to answer ceftain questiorrs, or stop participating at any time, andyou can ask me to stop recording at any time. There are no known risks associated with this projectwhich are greater than those ordir-rarily encountered in daily life, and nothing that affects you wouldresult if you decide not to parlicipate in whole or in part.
The records of tli is study will be kept private. Any written or publicly presented results will discussgroup findings, not information tliat would identifu you. All research records will be stored permanentlyin a locked file cabinet at Oklahorna State University and only researchers and individuals responsiblefor research oversight will have access to them. It is possible that the consent process and data collectionwill be observed by research oversight staff responsible for safeguarding the rights and wellbeing ofpeople who participate in research. We may play your recorded voice and show your image as part ofour repofts at academic conferences, on academic websites, or in acadernic videos. If at any time afteryou have completed this interview, you would like to withdraw, we will destroy your recording.
Ifyou have concerns or questions about thisstudy, such as scientific issues, how to do anypart of it, or to reporl an injury, please contactDennis R. Preston, Oklahoma State University40 5 -5 64 -063 6 or dennis. [email protected]
Ifyou have questions about your rights as aresearch volunteer, you may contact Dr. SheliaKennison, IRB Chair, 219 Cordel l North,Stillwater, OK 1 407 8, 405-317 -331 1 [email protected].
I have read and fully understand this form. I sign it freely and volLurtarily. A copy has been given to me.
Signature of Participant Date
I certifli that I have personally explained this document before requesting that the participant sign it.
Sisnature of Researcher
Please also sign below if you agree to the special case of being video-recorded. Please remember thatif you agree to this we may show your image as part of our repofts at academic conferences, on academicwebsites, or in academic videos.
Date
0Ih. ffie Univ.IRB
^wM-o/f//a.;rrlra .r. /g /12W8 A<oQ 67 .
S ignature of Partici pant Date
76
References
Atwood, E. Bagby. 1962. The regional vocabulary of Texas Austin: University of Texas Press.
Bailey, Guy, Jan Tillery & Tom Wikle. 1995. Reversal of near-merger. Paper presented at the
NWAV 24, Philadelphia : University of Philadelphia.
—. 1997. Methodology of a survey of Oklahoma dialects. SECOL (Southeast Conference on
Linguistics) Review 21.1-30.
Bailey, Guy, Tom Wikle, Jan Tillery & Lori Sand. 1993. Some patterns of linguistic diffusion.
Language Variation and Change 5.359-90.
Bakos, Jon. 2013. A Comparison of the Speech Patterns and Dialect Attitudes of Oklahoma:
Oklahoma State University Dissertation.
Britain, David. 1997. Dialect contact, focusing and phonological rule complexity: the
koineisation of Fenland English. Penn Working Papers in Linguistics: A Selection of
Papers from NWAV 25 4.141-69.
Brown, Vivian. 1990. The social and linguistic history of a merger: /ɪ/ and /ɛ/ before nasals in
Southern American English: Texas A&M University Dissertation.
—. 1991. Evolution of the merger of /I/ and /e/ before nasals in Tennessee. American Speech
66.303-15.
Feagin, Crawford. 1987. A closer look at the Southern drawl: Variation taken to exteremes.
Variation in Language, ed. by K. Denning, S. Inkelas, F. McNair-Knox & J. Rickford,
137-50. Stanford Univeristy: Department of Linguistics.
Feng, Gang & Eric Castelli. 1996. Some acoustic features of nasal and nasalized vowels: A
target for vowel nasalization. Journal of the Acoustical Society of America 99.3694-706.
77
Hawkins, Sarah & Kenneth N. Stevens. 1985. Acoustic and perceptual correlates of the non-
nasal--nasal distinction for vowels. Journal of the Acoustical Society of America
77.1560-75.
Herold, Ruth. 1990. Mechanisms of merger: The implementation and distribution of the low
back merger in eastern Pennsylvania. United States -- Pennsylvania: University of
Pennsylvania 9026574.
—. 1997. Solving the Actuation Problem: Merger and Immigration in Eastern Pennsylvania.
Language Variation and Change 9.165-89.
Labov, William. 1963. The social motivation of a sound change. Word 19.273-309.
—. 1972. Sociolinguistic Patterns Philadelphia: University of Pennsylvania Press.
—. 1994. Principles of linguistic change Cambridge [Mass.]: Blackwell.
—. 2006. The Social Stratification of English in New York City Cambridge: Cambridge
University Press.
—. 2010. Principles of Linguistic Change: Cognitive and Cultural Factors West Sussex: Wiley-
Blackwell.
Labov, William, Sharon Ash & Charles Boberg. 2006. The Atlas of North American English:
Phonetics, phonology and sound change Berlin: Mouton/de Gruyter.
Labov, William, Mark Karen & Corey Miller. 1991. Near-Mergers and the Suspension of
Phonemic Contrast. Language Variation and Change 3.33-74.
Liljencrants, Johan & Björn Lindblom. 1972. Numerical simulation of vowel quality systems:
The role of perceptual contrast. Language 48.839-62.
Maddieson, Ian. 1984. Patterns of Sounds Cambridge: Cambridge University Press.
78
Moulton, William. 1962. Dialect Geography and the concept of phonological space. Word 18.23-
32.
Peterson, Gordon E. & Harold L. Barney. 1952. Control methods used in a study of the vowels.
Journal of the Acoustical Society of America 24.175-84.
Preston, Dennis R. 1996. Where the Worst English is Spoken. Focus on the USA, ed. by E.W.
Schneider, 297-360. Amsterdam: Benjamins.
Roark, Michael Owen. 1979. Oklahoma Territory: Frontier development, migration, and culture
areas: Syracuse University Dissertation.
Rooney, John F., Wilbur Zelinsky, Dean R. Louder & John D. Vitek. 1982. This Remarkable
Continent: An Atlas of United States and Canadian Society and Culture College Station
[Tex.]: Published for The Society for the North American Cultural Survey by Texas A &
M University Press.
Solé, Maria-Josep. 2007. Controlled and mechanical properties in speech: a review of the
literature. Experimental approaches to phonology, ed. by M.-J. Solé, P. Beddor & M.
Ohala, 302-21. Oxford: Oxford University Press.
Southard, Bruce. 1993. Elements of Midwestern speech in Oklahoma. “Heartland” English, ed.
by T.C. Frazer, 229-43. Tuscaloosa and London: University of Alabama Press.
Thomas, Erik R. 2011. Sociophonetics: An introduction Hampshire: Palgrave Macmillan.
Tillery, Jan. 1997. The role of social processes in language variation and change. Language
Variety in the South, ed. by C. Bernstein, T. Nunnally & R. Sabino, 434-46. Tuscaloosa:
University of Alabama Press.
79
Trudgill, Peter & Tina Foxcroft. 1978. On the sociolinguistics of vocalic mergers: Transfer and
accommodation in East Anglia. Sociolinguistic Patterns in British English, ed. by P.
Trudgill, 69-79. London: Edward Arnold.