The status of the open-syllable constraint is severely weakened. The short- a of interna-tionally is clearly tense, and that of ceramic is in an intermediate position. Canadaand

 
Catholic, however, are clearly in the lax set.
As further evidence of the weakness of the open-syllable constraint in New Orleans,
one may consider the speech of Dr. John (Mac Rebennack), a prominent representative
of the New Orleans musical tradition who grew upin the Third Ward of the city at
mid-century. In a broadcast of March 16, 2005, Dr. John showed the following pattern
of tense and lax short-a.
27
Tense (closed syllable) answer , fancy, hand, bad, dad
Tense (open syllable) piano (2), classical , daddy, fascinate (2), Manny
Lax (closed syllable) that , cats , fact , that’s, at
Lax (open syllable) Allen
27
This broadcast is currently available at http://www.amroutes.com/programs/shows/20050316.html.
TRANSMISSION AND DIFFUSION 367
Dr. John’s tensing pattern includes vowels before nasals, voiced stops, and voiceless
fricatives, as in New York City, but short- a in open syllables is treated in the same
way as in closed syllables.
In New Orleans, as in Cincinnati, the local pattern is receding. Two other New
Orleans speakers analyzed acoustically are thirty-eight and forty-four years old; both
show the nasal short-a system, as in other Louisiana cities, Shreveport and Baton Rouge.
The history of New Orleans points to repeated and extensive connections with New
York City. While Cincinnati was an industrial rival of New York in the middle of the
nineteenth century, the city of New Orleans had intimate and complementary relations,
as the port of shipment for the cotton trade financed by New York bankers. This aspect
of the history of New Orleans is described by McNabb and Made ` re (1983:Ch. 3:1).
From 1803 until 1861, New Orleans’ population increased from 8,000 to nearly 170,000 . . . By 1830,
New Orleans was America’s third largest city, behind New York and Baltimore . . . During the Pre-Civil War period, a scarcity of capital in New Orleans forced seekers of large-scale investment to look
to New York, London, or Paris.
Berger summarizes the evidence for close relations between New Orleans and NYC
in the middle of the nineteenth century.
In the ante-bellum period, roughly between 1820 and 1860, financial, commercial and social relations
between the city and the South were at fever pitch: New York banks underwrote the plantation economy,
cotton was shipped routinely from New Orleans, Charleston, Savannah and Mobile to be trans-shipped
to England, and Southern planters regularly combined business with pleasure in the Big Apple of the
1800s. (1980:137)
Berger also cites the judgment of Foner 1941 as to the predominance of New York
City in New Orleans: ‘Down to the outbreak of the Civil War, New York dominated
every single phase of the cotton trade from plantation to market’ (1980:137).
Berger’s aim was to buttress the case for the derivation of the NYC palatalized mid
central vowel from New Orleans; this is the opposite direction of influence from the
one proposed here for the short-a pattern.
28
The gravity model and the historical facts
argue for a greater direction of influence from the larger city. Many descriptions of
commercial and social relations between New Orleans and New York are found in the
five-volume history The old merchants of New York City(Scoville 1863); the typical
pattern involves movement of New Yorkers to New Orleans, as can be seen in the
following examples. In Ch. 3 we read that Walter Barrett took a letter of credit for one
million dollars to New Orleans by way of Wheeling, hoping to outstrip his competitors
in buying upthat year’s cotton crop(p. 26). It is reported that the founder of the great
New York mercantile firm of E. K. Collins & Son had a house in New Orleans (p.
141). Among the oldest commercial firms of New York City was Brown Brothers &
Co., who established in 1842 a branch in New Orleans under the name of Samuel
Nicholson, ‘who had been many years their clerk’ (p. 187). Bradish Johnson, head of
the firm of Johnson & Lazarus, had a brother Henry who was located on a plantation
in New Orleans. When Henry died he left the plantation to Bradish, who proceeded to
New Orleans and established more favorable conditions for the 250 slaves, many of
whom were able to purchase their own freedom (p. 185). In the description of the
prominent Seixas merchant clan, founded by Benjamin Seixas in 1780, we read: ‘Madi-28
Both directions are of course possible, and it is plausible that palatalization of work, third , and so forth
is derived from the South, as PEAS (Kurath & McDavid 1961) shows that it is widely used in several
Southern areas.
LANGUAGE, VOLUME 83, NUMBER 2 (2007) 368
son [Seixas] is in New Orleans, and a partner in the large firm of Glidden and Seixas’
(Vol. 2, p. 127).
Among the bankers closely related to New Orleans were many representatives of the
large Sephardic Jewish families (Lazarus, Seixas). Scoville underlines the importance of
the Jews in many places: ‘The Israelite merchants were few then [1790], but now? they
have increased in this city beyond any comparison. There are 80,000 Israelites in the
city. It is the high standard of excellence of the old Israelite merchants of 1800 that
has made this race occupy the proud position it does now in this city’ (p. 127). We
can see how intimate the relations were between the Jewish population of the two cities
by examining Korn’s history The early Jews of New Orleans (1969), which deals with
social and business relations from 1718 to 1812. References to New York City are
found on fifty-five pages, more than any other city.
29
Following the publication of ANAE, I received a letter from Mr. Herman S. Kohl-meyer, Jr., Senior Vice President of the investment firm A. G. Edwards, who described
himself as ‘the last person in New Orleans who still makes his living from the cotton
trade’. His account leaves no doubt that Jewish merchants with strong New York City
connections played a formative role in the upper class speech of New Orleans.
I am the great-grandson of some of our topcotton merchants . . . as is my closest friend. They were
all German Jewish immigrants who came over in the 1830–1860 era ….Iremember very well friends
of my father’s generation who talked about how hard they ‘woiked’ before they went home to their
house on ‘Foist’ Street. That was very much our upper class speech, as much with the Christians and
with the Jews.
30
The detailed linguistic resemblances between New York City and New Orleans in-volve both of the pivot points that have been found to determine the main directions
of development of North American dialects: the status of short- o as an integral phoneme
distinct from long open-o and the status of short- a (Labov 1991). As in New York,
the New Orleans raised /oh/ ensures the separate status of short- o as the phoneme
/o/.
31
As in New York, New Orleans divides short-a into two distinct classes, with the
tense vowel occurring before front nasals, voiced stops, and fricatives in closed sylla-bles, and the lax vowel before voiceless stops and liquids. The New Orleans adaptation
is only superficially similar to the NYC configuration, however: it is a phonetically
conditioned set of allophones rather than a grammatically and lexically specified distri-bution.
In the four cases of diffusion of the NYC short- a pattern presented above, phonetic
conditioning by the following segment is the common thread, though the phonetic
pattern is not perfectly transmitted. The voiced velars are excepted from the voiced
stops, and tensing before voiceless fricatives is sometimes generalized to voiced frica-tives. But the most regular differences are found at a more abstract level. The function-word constraint is lost: with few exceptions,can, am, and, have , has, had are tense,
29
Korn’s book refers to Charleston on forty-three pages, Savannah on five, and Boston on six.
30
Mr. Kohlmeyer referred to an oral tradition in his family that the New York City influence in New
Orleans was from a single teacher from Brooklyn who arrived in the 1890s. Marc Caplan of New Orleans
told me of an oral tradition in his family that attributed New York City influence to the period late in the
nineteenth century when New Orleans docks were rebuilt with the helpof large numbers of laborers from
New York City. I have found no written evidence for this.
31
The influence of the Jewish community, detailed above in the historical data, appears phonetically in
the raised /oh/ of New Orleans. There is a marked tendency for second and following generations of Jews
to raise this vowel to upper mid and lower high position, more so than other ethnic groups: see Labov 1966
for New York City and Laferriere 1979 for Boston.
TRANSMISSION AND DIFFUSION 369
though they are always lax in NYC. The second major difference is the loss of the
constraint against tensing in open syllables, quite general though not complete in New
Orleans. It might seem at first glance that this represents the loss of a phonological
constraint. But on reflection it may be seen as the loss of the effect of inflectional
boundaries in closing the syllable. When short-a is tensed in all open syllables, there
is no longer a difference between [Cardinal ]/m +ni ?/ and /m +hn#i ?/[ the pumps], or
between monomorphemic /b +nUr/ and /b +hn#Ur/, a person who bans. The adults who
adopted the NYC system did not observe that tense /m+hn#i ?/, /b +hn#Ur/, /p +hs#i ?/,
/p +hs#Ur/ are bimorphemic, while /m +ni ?/, /b +nUr/, /k +sUl/, /b +fUl/ are not. Accord-ingly, they generalized the tensing in bimorphemic words to all words of this phonetic
shape. This is consistent with the proposition that the main agents in diffusion are
adults who are less likely to observe and replicate abstract features of language structure.
3.5. D IFFUSION ACROSS COMMUNAL GROUPS . The discussion so far has concerned the
diffusion of linguistic structures from place to place. The speech communities described
so far—New York, Albany, Cincinnati, New Orleans—are formed by the population
defined in American society as the white mainstream. They are geographical unities,
differentiated internally by social class, but separated sharply from the African Ameri-can and Latino populations in the same cities. Most American cities include three major
communal groups, in the sense defined in Blanc’s 1964 study of the Muslim, Christian,
and Jewish dialects of Baghdad. Contacts between such communal groups are primarily
among adults, and when linguistic patterns diffuse from one group to the other the
same loss of structure that was observed in geographic diffusion can be expected.
This is a major topic to be explored in relation to the many studies of African
American and Latino dialects in the United States. One example can be cited here,
from Henderson’s 1996 study of short- a in the African American community of Phila-delphia. As indicated above, the Philadelphia short-a distribution into tense and lax
classes is similar to that of New York City in having grammatical conditioning and
the open-syllable constraint. Among the features in which it differs are (i) short-a is
tense before voiced stops in only three words (mad, bad, glad ), and (ii) short- a is lax
before nasals in irregular verbs ( ran, swam, began ).
32
Table 1 compares short- a tensing
in the spontaneous speech of one hundred white Philadelphians reported in Labov
1989b with the thirty speakers of Henderson’s study.
EURO -AMERICANS AFRICAN AMERICANS
(Labov 1989) (Henderson 1996)
% tense % tense
Following segment
Normally tense in white Philadelphia dialect
before nasal coda 96 95
before voiceless fricatives 98 69
mad, bad, glad 99 83
Normally lax in white Philadelphia dialect
before intervocalic nasals 1 43
ran, swam, began 19 71
TABLE 1. Tensing of short-a for Whites and African Americans in Philadelphia.
For the normally tense classes, the white Philadelphians have the tense vowel nearly
100% of the time. African Americans tense equally consistently before nasals and come
32
And wan, the vernacular preterite of win.
LANGUAGE, VOLUME 83, NUMBER 2 (2007) 370
close for the mad, bad, glad subclass, but show less tensing of short-a before voiceless
fricatives in path , bath , pass , and so forth, with only 69% tense. Although there is
some lexical diffusion in open syllables, white Philadelphians show only 0.4% tensing
overall before intervocalic nasals. The open-syllable constraint is much weaker among
African Americans; almost half of the tokens in this environment are tense. Finally,
one can observe that the grammatical constraint that laxes short- a in irregular verbs
ending in nasals is almost missing in the diffusion to the African American community:
only 29% are lax as compared with 80.7% lax among white Philadelphians. This loss of
structural detail in diffusion across communal groups echoes the patterns of geographic
diffusion; in both situations, contact is largely through adult speakers.
3.6. THE TRANSMISSION AND DIFFUSION OF MERGERS AND SPLITS. The argument so far
has not considered the type of structural diffusion that is most frequent and most promi-nent in historical linguistics and dialectology: mergers. Herzog’s corollary of Garde’s
principle (Herzog 1965, Labov 1994) states that mergers expand geographically at the
expense of distinctions; there is massive empirical evidence of such expansion.
33
Though the adoption of a merger is not conventionally considered to be structural
borrowing, it must be considered so, since the recipient dialect loses one of its categories
in adopting the structure of the expanding dialect. Up to this point, I have been arguing
that adults do not easily acquire new structural categories, but the evidence does not
so far bear on the loss of a category.
Herold’s (1990) proposal for the diffusion of a merger is that speakers of a two-phoneme system in contact with a one-phoneme system find that the contrast is not
useful and so cease to attend to it. There is ample evidence that merger in perception
precedes merger in production (Di Paolo 1988, ANAE Ch. 9), and near-mergers give
us a static view of such a situation (Labov 1994:Ch. 12, Labov et al. 1991). But this
does not tell us how a merger in the speaker’s perception is transmitted to the speaker’s
children. There are indeed numerous cases of a contrast strongly maintained among
adults but solidly merged in the speech of their children, but the mechanism of such
transmission is still obscure.
34
It is possible that adults come to lose the distinction in
production as well. However, none of the real-time panel studies—restudies of the
same individuals over their life span—have dealt with ongoing mergers that would
produce evidence of adults collapsing phonemic categories during their lifetimes (Ced-ergren 1988, Trudgill 1988, Sankoff 2001).
Until more evidence on the diffusion of mergers is acquired, my discussion of limita-tions on adult language learning must be focused on the acquisition of new grammatical
constraints. In rule-based generative systems, this may refer to the acquisition of a rule
that operates within the phonological cycle. In constraint-based systems, it means raising
the ranking of a grammatically defined constraint over the ranking of a phonetically
defined constraint.
33
ANAE Chapter 8 shows that the distinctions between /hw/ and /w/, /ohr/ and / :hr/, /iw/ and /uw/ have
all but disappeared in the United States, although they were strongly maintained in both the North and the
South in the records of the mid-twentieth century (Kurath & McDavid 1961). The low back merger of /o/
and /oh/ has expanded in some areas with comparable speed. The Philadelphia LVC project interviewed
adolescents at a Pottsville recreational park in 1977. When Herold (1990) returned to the same site eleven
years later, she found that the percent of those judging cot and caught as ‘the same’ jumped from 17 to
100% for girls, and from 29 to 67% for boys.
34
Herold 1990 provides acoustic analyses of adult speakers with a stable distinction of the low back
vowels /o/ and /oh/ and their children with complete merger.
TRANSMISSION AND DIFFUSION 371
The continuity of the NYC short-a system from 1896 to the present and the uniformity
of the Mid-Atlantic short- a system in Philadelphia, Reading, Wilmington, and Balti-more all indicate that such patterns can be faithfully transmitted across generations
through children’s language learning abilities. There is evidence, however, that a pattern
of this complexity cannot be learned as a second dialect, even by children. Payne studied
the acquisition of the Philadelphia dialect by children of out-of-state parents in King
of Prussia (1976, 1980). She found that children under ten years of age acquired the
phonetic variables of the Philadelphia system after only a few years in King of Prussia,
but only one of thirty-four children of out-of-state parents acquired the lexical and
grammatical conditioning of the short- a system. There were, however, differences in
the degrees of approximation, depending on the parents’ dialect (Payne 1976, Labov
1994). Children of New York City parents approximated the Philadelphia system much
better in identifyingmad, bad, glad as the only lexical items in which short-a is tense
before voiced stops than in acquiring the general Philadelphia rule that short-a is always
lax before back consonants ( cash , rash , smash , etc.). Children of parents from regions
with allophonic short-a distributions of types (i–iii) above showed the opposite bias,
favoring phonetic generalization. This strongly suggests that the NYC families had
acquired their own short- a pattern as a lexical list rather than as a rule-governed distribu-tion. If we conclude that the NYC short- a distribution is a phonemic split, it does not
lead to the conclusion that it diffuses as a phonemic split. What we have seen in North
Plainfield, Albany, Cincinnati, and New Orleans is the diffusion of a close approxima-tion to the segmental conditioning of the NYC system, without its lexical, grammatical,
or syllabic conditioning. In other words, adults with allophonic short- a systems of types
(i–iv) will approximate the NYC pattern as a rule or constraint system of the type they
have acquired as children.
This conclusion is consistent with the fact that the distinction between transmission
and diffusion is maximal in the case of splits. The converse of Garde’s principle is that
splits are rarely reversed. Britain’s (1997) account of the complexities of the /u– R/ split
in the Fens shows the irregular result of a rare case of expansion of the split where the
two-phoneme system is favored by social prestige. The constraint on learning a new
phonemic contrast applies equally to studies of the children of immigrant parents.
Trudgill examined the ability of twenty adults born in Norwich to reproduce the local
distinction between the vowel classes ofown [Run] andgoal [gu:l]. Ten whose parents
were born in Norwich did so; the ten whose parents were born elsewhere could not
(Trudgill 1986:35–36).
It is apparent that an UNBROKEN sequence of parent-to-child transmission is required
to maintain complex patterns of phonetic, grammatical, and lexical specification like
the NYC short-a pattern. Therefore, if speakers from other dialect areas enter the com-munity in large numbers, their children will dilute the uniformity of the original pattern.
Although the Mid-Atlantic dialects are quite stable, there is some indication of such a
weakening. Lexical diffusion of tensing of short-a in open-syllables before /n/ has been
traced since 1980 (Labov 1989b, Roberts & Labov 1995); some neighborhoods report
general tensing before /l/ (Banuazizi & Lipson 1998); still other neighborhoods show
shifting to the default nasal system, as in certain small towns of southern New Jersey
(Ash 2002).
To examine more closely the difference between transmission by children and diffu-sion by adults, I turn to a complex system that is free of such lexical and grammatical
specification, the Northern Cities Shift. The structural complexity involved here has
LANGUAGE, VOLUME 83, NUMBER 2 (2007) 372
to do with the intricate interrelations of vowels as they evolve in chain shifts within
and across subsystems (Martinet 1955, Moulton 1960).
4. DIFFUSION OF THE NORTHERN CITIES SHIFT . The NORTHERN CITIES SHIFT(NCS) is
the rotation of six vowels shown in Figure 14 (Labov et al. 1972, Eckert 1988, 1999,
Labov 1991,ANAE Ch 14, Gordon 2001). The NCS was initiated by the general tensing
and raising of short-a to mid and high position. The absence of vowel tokens in low
front position then led to a shift of two neighboring vowel classes into that vacant
space: short-o shifted frontward and short- e shifted downward. This was followed by
the lowering and fronting of long open- o. In later developments, short- e shifted back
toward / R/, and /R/ moved back to the position formerly occupied by long open- o
(/oh/), while /i/ moved down and back. The NCS has developed incrementally in all
cities of the Inland North, including Syracuse, Rochester, Buffalo, Cleveland, Toledo,
Detroit, Flint, Grand Rapids, Kalamazoo, Gary, Chicago, Kenosha, Milwaukee, and
Madison. The most remarkable fact about the NCS is its uniform distribution across
the vast area surrounding the Great Lakes ( ANAE Chs. 11, 14).
bit
bet
cat
cot
caught
but
i
e
æ o
oh
1
2
3
5 4
6
V
FIGURE 14. The Northern Cities Shift.
Figure 15 shows how the NCS is realized in the vowel system of Kitty R. of Chicago
when she was interviewed at the age of fifty-six in 1993. The general raising of / +/
to upper mid position is shown by the solid black squares, and the fronting of /o/ by
the small empty squares with five tokens well front of center. Diamonds indicate the
backing of /e/ with mean F2 of 1864 Hz, only 320 Hz higher than the mean F2 of /o/
(1544 Hz). Wedge is shifted well to the back, overlapping /oh/, which has not lowered
extensively.
Figure 16 displays the geographic distribution of the Northern Cities Shift. Since
the NCS involves a complex rotation of its elements, the measurement of any one
vowel tells us little about the progress of the shift. ANAE relies on structural relations
among NCS vowels to mapthe progress of the shift. One such relation is the combined
effect of stages 2 and 4 of the NCS (Fig. 14), measuring the extent to which the backing
of short /e/ in bet, dead , and so forth is accompanied by the fronting of short /o/ in
cot, odd, and so forth. For most North American dialects, /e/ is a front vowel and /o/
is a back vowel, with mean differences in F2 of about 1000 Hz. For speakers most
fully engaged in the NCS, /e/ is close to or aligned with /o/ along the front-back
dimension. In Fig. 16, the gray circles indicate speakers who satisfy the ED CRITERION,
for whom the difference between the mean F2 of /e/ and the mean F2 of /o/ is less than
375 Hz. The Inland North—the region in which the NCS is operating—is delineated by
this measure.
TRANSMISSION AND DIFFUSION 373
FIGURE 15. NCS in vowel system of Kitty R., 56 (1993), Chicago, IL, TS 66.
The earliest records we have of the chain shift of /+/, /o/, and /oh/ date from the
1960s. Yet there is reason to think that the initiating event of the NCS took place a
hundred years earlier with the construction of the Erie Canal in western New York
State. A koine´ ization of various complex short- a systems to the simple general tensing
seems to have occurred when workers and migrants from all over the northeast were
integrated into the rapidly expanding cities of Rochester, Syracuse, and Buffalo (Labov
2004). The unrounding and centralization of /o/ had already taken place in western
New England (ANAE Ch. 16). With westward migration of entire communities, the
conditions for the chain shift were transmitted faithfully across the Inland North as far
as Wisconsin.
The linguistic boundary separating the Inland North from Midland vowel patterns
is the sharpest and deepest division in North American phonology. The isogloss bundle
that separates these two areas combines six measures of the progress of the NCS, the
southern limit of Canadian raising of /ay/, and the southern limit of dialects with /aw/
backer than /ay/ (ANAE Ch. 11). Figure 16 shows that the front-back approximation
of /e/ and /o/ is generally absent in the Midland region, except for St. Louis and nearby
communities. The city of St. Louis, located squarely in Midland territory, has recently
developed many of the elements of the NCS. This city has long been known to display
a mixture of Northern, Midland, and Southern features (Murray 1993, 2002), but recent
decades have witnessed a strong shift to Northern phonology. The characteristic St.
Louis merger of /ahr/ and /:hr/ in are and or , card and cord , barn and born , and so
LANGUAGE, VOLUME 83, NUMBER 2 (2007) 374
FIGURE 16. The ED criterion of advance of NCS in Inland North and St. Louis corridor.
Gray symbols ?F2(e)?F2(o)? 375 Hz.
forth, has all but disappeared among younger speakers, who display instead the general
merger of or and ore, cord and cored, along with a clear separation of this class from
/ahr/ in are and card .
Figure 17 shows both the modern St. Louis vowel pattern and the traditional merger
before /r/ in the system of Marvin H., a manufacturer’s wholesale representative of
German background who was forty-eight when interviewed in 1994. At the upper right,
one can see tightly clustered the traditional /ohr/ class— hoarse , four , Ford.Inmid
position is the class of /:hr/—for, born , horse, corn , morning—alongside /ahr/ in part ,
far, and barn . The distinction of hoarse and horse, four and for is well illustrated, as
well as the identity offar and for, born and barn . At the same time, the distribution
of the NCS vowels matches the Chicago pattern of Fig. 15 quite well. All / +/ are raised
to mid position, /o/ is well fronted, and /e/ is backed close to the midline. The difference
between the second formants of /e/ and /o/ is only 134 Hz. Wedge is moderately back
and some tokens of /oh/ are quite low. It is apparent that Marvin H. has combined the
traditional St. Louis pattern with the Northern Cities Shift.
This recent development in St. Louis is not an independent phenomenon, distinct
from the chain shift in the Inland North.
35
ManyANAE maps show diffusion of NCS
features along a narrow corridor extending from Chicago to St. Louis along Route
35
The NCS operating in the Inland North is here assumed to be governed by the mechanical operation
of probability-matching by language learners, as described in Labov 2001:Ch. 20. The outcome takes the
form of pull shifts and push shifts as described in Martinet 1955, but without any purposive intent to preserve
contrasts.
TRANSMISSION AND DIFFUSION 375
FIGURE 17. NCS and merger of /:hr/ and /ahr/ for Martin H., 48 (1994), St. Louis, MO, TS 111.
On this chart: /Ohr/?/:hr/.
I-55 (Figure 18).
36
The ANAE data for this corridor is based on speakers from three
cities along the interstate highway (Fairbury, Bloomington, Springfield), along with
four speakers from St. Louis.
37
In Fig. 16, fifty-nine of the sixty-seven speakers within the isogloss satisfy the ED
criterion, a HOMOGENEITY of .88. A similar proportion of speakers in the St. Louis
corridor do so—seven out of nine.
A second measure, displayed in Figure 19, shows even more clearly how the St.
Louis corridor is differentiated from its Midland neighbors. Stage 2 of the NCS, the
fronting of /o/, and stage 5, the backing of / R/, has the effect of reversing the relative
front-back positions of these two vowels as compared with neighboring dialects. The
UD CRITERION used by ANAE to define the progress of the NCS defines the speakers
involved in this chain shift as those for whom /R/ is further back than /o/ (gray circles
on Fig. 19). Of all of the measures of the progress of the NCS, this yields the sharpest
differentiation between the Inland North and the Midland. Homogeneity within the
Inland North is even greater than for the ED measure: sixty-five out of sixty-seven
subjects in the Inland North satisfy the UD criterion, or .94. The almost total absence
of gray symbols in the Midland area of Fig. 15 contrasts with the five gray symbols
in the St. Louis corridor. Though this corridor is represented in ANAE by only four
cities and nine speakers, the probability of the occurrence of this feature in the corridor
36
The interstate highway I-55, built just after World War II, is now the main route for Chicago-St. Louis
travel, but it follows the path of earlier traffic, in particular the Illinois-Central Railroad, which was built
in 1856 to connect Cairo in the southern tipof Illinois with Galena and Chicago.
37
The city of Peoria is not far from I-55, but it is not on the direct route.
LANGUAGE, VOLUME 83, NUMBER 2 (2007) 376
FIGURE 18. Corridor along Route I-55 from St. Louis to Chicago.
by chance is less than one out of a thousand.
38
It is, however, significantly less frequent
than in the Inland North.
39
Figures 16 and 19 illustrate the diffusion of the NCS along I-55 from Chicago to
St. Louis. It appears, however, that the NCS along this corridor is not the same linguistic
phenomenon as in the Inland North itself; there is reason to believe that the systematic
chain-shift mechanism, triggered by the general raising of short- a, is not driving the
shift in the St. Louis corridor.
Figure 20 is a map of the same region displaying speakers for whom the NCS is
complete—who show all relevant criteria. In addition to the ED and UD criteria, we
have the following.
38
The Midland distribution is 75 to 1, but since the null hypothesis for the nine tokens within the corridor
would have fewer than five tokens in a cell, Fisher’s exact test is appropriate, yielding p ?0.00026.
39
The difference in homogeneity between the St. Louis corridor and the Inland North has a probability
of 0.0017 by Fisher’s exact test.
TRANSMISSION AND DIFFUSION 377
FIGURE 19. The UD criterion of advance of NCS in Inland North and St. Louis corridor. Gray symbols ?
F2(R) ? F2(o). Solid isogloss ?Inland North as defined by ED measure.
FIGURE 20. Speakers who show all criteria of NCS: AE1, O2, EQ, ED, and UD. Solid isogloss?Inland
North as defined by ED measure.
LANGUAGE, VOLUME 83, NUMBER 2 (2007) 378
AE1: general raising of /+/ in nonnasal environments, F1( +) ? 700 Hz
O2: fronting of /o/ to center, F2(o)? 1500 Hz
EQ: reversal of the relative height and fronting of /e/ and /+/: F1(e) ? F1(+)
and F2(e) ? F2(+)
Figure 20 shows that twenty-eight of the sixty-seven Inland North speakers meet
this strict criterion—42%. Sixteen of the twenty-eight are in the largest cities: Detroit,
Rochester, Syracuse, Chicago. By contrast, the St. Louis corridor shows only one such
speaker—Martin H. of Fig. 17—and no one else outside of the Inland North.<br