LINGUISTIC CHANGE IN NORTH AMERICA

Recent studies of linguistic change in North America show that it flows in two different directions. On the one hand, the white mainstream dialects have invested in radically different vowel systems. The Atlas of North American English (Labov, Ash, & Boberg, Reference Labov, Ash and Boberg2006) finds steadily increasing regional differentiation (Inland North, Canada, the Middle Atlantic States, western Pennsylvania), a calm maintenance of traditional patterns (Eastern New England, New York City, North Central, the Midland), or a slowly receding regional generalization (the South). Several moderate-sized cities have abandoned their local configurations but have replaced them with a neighboring regional pattern—Charleston (Baranowski, Reference Baranowski2007), Cincinnati (Boberg & Strassel, Reference Boberg and Strassel2000), and St. Louis (Labov, Reference Labov2007).

We find a very different situation when we turn to African American Vernacular English (AAVE). This dialect has undergone rapid grammatical evolution on a national scale with no detectable regional differentiation. Remarkably similar findings have appeared in studies of morphosyntactic features (absence of verbal –s, absence of possessive attributive –s, presence of plural –s, variable deletion of copula and auxiliary) in cities throughout the country (New York [Labov, Cohen, Robins, & Lewis, Reference Labov, Cohen, Robins and Lewis1968]; Washington, DC [Fasold, Reference Fasold1972]; Detroit [Wolfram, Reference Wolfram1969]; San Francisco Bay area [Mitchell-Kernan, Reference Mitchell-Kernan1969]; Philadelphia [Cofer, Reference Cofer1972; Labov & Harris, Reference Labov, Harris and Sankoff1986]; Los Angeles [Baugh, Reference Baugh1983]). Newly incrementing elements of the tense and aspect system (habitual be, preterite had) have appeared simultaneously in widely separated cities across the United States (Cukor-Avila & Bailey, Reference Cukor-Avila, Bailey, Arnold, Blake and Davidson1996; Labov et al., Reference Labov, Cohen, Robins and Lewis1968; Rickford, Ball, Blake, Jackson, & Martin, Reference Rickford, Ball, Blake, Jackson and Martin1991).

Given this uniformity, nine researchers on AAVE were able to jointly issue a Summary Statement describing the common features that educators would have to take into account (Labov, Reference Labov2013:Appendix). This national distribution appears to be largely a product of the 20th century, the outcome of the Great Migration of southern blacks to northern cities beginning in 1914 (Cukor-Avila & Bailey, Reference Cukor-Avila, Bailey, Arnold, Blake and Davidson1996).

A somewhat different situation has begun to emerge from studies of AAVE phonology, where considerable regional differentiation is to be found. The Summary Statement describes several common phonological features (th-fronting; merger of /i/ and /e/ before nasals; /l/ vocalization), but one feature was reported as “regional”: the vocalization of /r/. Studies of AAVE in New York City found that adolescent African American speakers had 98% to 100% vocalization, considerably exceeding the white pattern (Labov et al., Reference Labov, Cohen, Robins and Lewis1968, I:99–106). In r-pronouncing Philadelphia, core speakers of AAVE showed as much as 71% r-vocalization (Myhill, Reference Myhill1988). The general pattern shows an influence of the surrounding r-pronouncing mainstream dialect on an originally r-less AAVE. Yet the most striking regional feature of AAVE is an increase in the influence of coda /r/ on the vowel: the St. Louis realization of front vowels /ihr/ and /ehr/ as mid-central [ɝ] (made nationally famous by the hip-hop artists Chingy in “Right Thurr” and Nelly in “Hot in Herre”). Blake and Shousterman (Reference Blake and Shousterman2010) track the development of this sound change within African American English in St. Louis and East St. Louis and trace its origins to Memphis.

In contrast, most other reports of regional differences in AAVE phonology show an approximation to the surrounding regional pattern of the white community. In Pittsburgh, Gooden and Eberhardt (Reference Gooden and Eberhardt2007) examined the use of well-known features of the vowel system by local black speakers. African Americans showed 7% of monophthongal /aw/ as in “dahntahn,” compared with 21% for whites. The low back merger was represented by the backing and rounding of /o/ in cot, Don, etc.; here, Whites showed 52% but Blacks only 22%. In Milwaukee, Purnell (Reference Purnell, Yaeger-Dror and Thomas2010) demonstrated that local African Americans shifted their vowel systems measurably in the direction of the white dialect in interaction with white speakers of that dialect. In Michigan, Preston and colleagues found that minority ethnic groups exhibit vowel shifting in the direction of the Northern Cities Shift (Evans, Ito, Jones, & Preston, Reference Evans, Ito, Jones, Preston, Murray and Simon2006). For African Americans, Jamila Jones's (Reference Jones2003) study showed a partial reflection of the Northern Cities Shift (Figure 1): short-a has moved up to approximate the height of short-e but does not go beyond it. This diagram may reflect a general tendency in African American phonology that will also emerge in the present study of African American phonology in Philadelphia.

Figure 1. Short front vowels in the African American vowel system in Detroit (Jones, Reference Jones2003), with short-a approximating the position of short-e.

AFRICAN AMERICAN PHONOLOGY IN PHILADELPHIA

The studies of African American phonology just cited show a regional effect: a measurable influence of the surrounding white dialect on the phonetic parameters of black speech. A more complete comparison of local black and white phonology can be obtained from the Philadelphia Neighborhood Corpus (PNC) by means of the Forced Alignment and Vowel Extraction (FAVE) computational analysis of vowel systems. Labov, Rosenfelder, and Fruehwald (Reference Labov, Rosenfelder and Fruehwald2013) (henceforth LRF) traced 100 years of sound change in the speech of 264 white adults in the PNC. LRF focused entirely on mainstream whites, who compose less than half of the Philadelphia total population of 1.55 million.Footnote ¹ There is little difference in the use of Philadelphia phonology among the whites of Irish, Italian, Jewish, Polish, Greek, and other European backgrounds. (Labov, Reference Labov2001:245–250). But African Americans are a different story. This study examines the vowel systems of 36 African American speakers in that corpus to determine how much they participate in the ongoing changes and the degree to which they conform to the traditional phonological pattern of the local white mainstream.

African American speakers in PNC are drawn from two sources:

1. 1. Twelve were from the project of the 1980s on “The Influence of Urban Minorities on Linguistic Change” (Ash & Myhill, Reference Ash, Myhill and Sankoff1986; Graff, Labov, & Harris, Reference Graff, Labov, Harris and Sankoff1986; Labov & Harris, Reference Labov, Harris and Sankoff1986). They show the most characteristic forms of the African American vernacular, as defined by Baugh (Reference Baugh1983): African Americans who live with, work with, and speak with other African Americans primarily. Speakers were recorded in intimate exchange with well-known interlocutors, where linguistic variables reach their most categorical form.

2. 2. The other 24 speakers were drawn from the neighborhood studies of LING560, The Study of the Speech Community, as for most of the PNC interviews. These were people living in racially mixed neighborhoods—Frankford, Port Richmond, Mount Airy, and West Philadelphia—where they had considerable interaction with White and Hispanic neighbors.

Figure 2 shows the distribution of the 36 African American speakers by date of birth and year of interview against the main body of white speakers in the PNC. African Americans are not as evenly distributed as one would like, but year of interview ranges from 1977 to 2002 and date of birth from 1910 to 1983; they are evenly divided among men and women.

Figure 2. Philadelphia Neighborhood Corpus subjects analyzed for the current study by race, date of birth, and year of interview.

PARTICIPATION OF AFRICAN AMERICANS IN PHILADELPHIA SOUND CHANGES

LRF reported two Philadelphia sound changes that progressed in a linear fashion across the century: (eyC),Footnote ² the raising along the front diagonal of /ey/ before a consonant in made, pain, etc. and (ay0), the centralization of /ay/ before voiceless consonants in right, wife, etc.Footnote ³ The entire white community participated in this raising, with women and college graduates slightly ahead in the most recent decades. These are not salient variables: (eyC) has never been cited in public as a feature of Philadelphia English and (ay0) rarely is. Figure 3 compares the mean values of these two variables for whites and African Americans by date of birth. The dashed lines are the locally weighted regressions for the white majority and the solid lines for African Americans.Footnote ⁴ The white majority of African Americans lag behind the white majority, though one outlier born in 1979 closes the gap. The (ay0) graph registers centralization as lowering of F1. Some tendency to follow the white majority appears among the older speakers, but among the younger African American speakers, only the same outlier approximates the white trend.

Figure 3. Locally weighted regression analyses of two nonsalient Philadelphia sound changes by date of birth and race. a = African American, n = 36; w = White, n = 330. Gray area = 95% confidence intervals.

LRF found unexpected reversals in the development of back upgliding vowels /aw, ow, uw/ in the white mainstream, which had shown a steady pattern of increased fronting in the 1970s. The reversal began with speakers born in the 1950s; though it appears at the same time in four variables, there is at present no explanation as to why it occurred at this time.Footnote ⁵ LRF points out that the variables that are reversed are those that are also found in Southern dialects, while those that continue in the same direction are found in Northern dialects. It is also true that the two sets differ in the degree of salience in the speech community. The strongly fronted forms of the back upgliding vowels show a certain degree of social awareness for whites, as demonstrated in subjective reaction experiments (Labov, Reference Labov2001:208–222).

Figure 4 shows that the reversal is not to be found in our African American subgroup. African Americans do not echo the mainstream rise and fall pattern for (aw) in south, out, down, etc., nor for checked (owC), in road, goat, phone, etc. Their movement, if any, appears to be in the opposite direction.

Figure 4. Locally weighted regression analyses of two back upgliding Philadelphia sound changes by date of birth and race. /aw/ in south, out, down, etc. /owC/ in road, goat, phone, etc. a = African American, n = 36; w = White, n = 330. Gray area = 95% confidence intervals.

The opposing shifts in the realization of (aw) and (ow) take on greater significance in the light of the experimental results of Graff et al. (Reference Graff, Labov, Harris and Sankoff1986). Sentences spoken by African Americans with characteristic back forms of /aw/ and /ow/ were resynthesized with fronted nuclei characteristic of white speakers.Footnote ⁶ These were incorporated into a matched guise experiment in which White, African American, and Hispanic subjects were asked to say whether the speakers were Black or White. The results showed that the controlled fronting of these vowels effectively converted judgments of ethnicity from Black to White in the great majority of judges.Footnote ⁷

A third pattern in the white mainstream dialects was found by LRF for the most salient sound changes, which remained at a high level for the first part of the century but showed a general decline in the second half. These are (æh), the product of raising and fronting along the front diagonal of a subset of short-a defined by intricate phonological, lexical, and grammatical conditions, and (oh), the corresponding back ingliding vowel in talk, lost, saw, etc. In the self-report tests of the 1970s, upper mid forms of (æh) were most often the subject of overt comment such as “Unfortunately, it's South Philly slang, not the best pronunciation,” and (æh) showed the greatest disparity between self-report and actual use (Labov, Reference Labov2001:203). In matched guise tests, upper mid forms of (æh) such as [e:ɔ] produced the greatest decline in job suitability (Labov, Reference Labov2001:213). Though (oh) was not included in these field experiments, it is the topic of increasing mention in public discourse on the Philadelphia dialect as the stereotype water with a high back vowel [wuɾɚ]).

Figure 5 presents the corresponding locally weighted regressions by date of birth and race for these most salient variables. There is little evidence of African Americans following the pattern of the white community; on the contrary, African Americans follow the inverse pattern, moving toward the traditional stigmatized target rather than away from it. The number of younger speakers born in the 1980s is too small to produce a reliable result, as the enlarged gray areas representing 95% confidence intervals indicate. But none of these figures show African Americans participating in the pattern of the white Philadelphia sound changes.

Figure 5. Locally weighted regression analyses of two ingliding Philadelphia sound changes by date of birth and race. /æh/ in man, bath, mad, etc., and /oh/ in water, talk, off, etc. a = African American, n = 36; w = White, n = 330. Gray area = 95% confidence intervals.

PARTICIPATION IN THE TRADITIONAL SHORT-A SPLIT

The phonetic path of the tense /æh/ class is a relatively superficial measure of African American participation in Philadelphia phonology. To obtain a more precise indication of the participation of African Americans in the system of Philadelphia phonology, we must consider the structural conditions that define the Philadelphia split between lax and tense short-a (Ferguson, Reference Ferguson and Smith1975; Labov, Reference Labov, Fasold and Schiffrin1989). The core set of conditions that require tense vowels are given as (1).

1. (1) In syllables closed with front nasals and voiceless fricatives /m, n, f, s, θ/:

   ham, hand, past, half, bath, etc.

   and in open syllables before inflectional suffixes: planning, passes, etc.

   and before /d/ in three affective adjectives: mad, bad, glad

   but not in irregular verbs: ran, swam, began

   and not in function words: am, an, and, can

   and not in polysyllabic words with zero codas: aspirin, athletic, etc.

   and not in learned words: alas, haft, etc.

The main focus of our examination is on the extent to which the African American speakers integrate these conditions into their phonology. Table 1 shows an early answer to this question in Henderson (Reference Henderson1996), a study of the short-a split among 30 middle-class African Americans in West Philadelphia. Closed syllables before nasals are close to categorically tense but short-a before voiceless fricatives shows only 69% tense, and the lexical set of three words before /d/ is only 83% tense as compared to the 99.6% for whites found in the 1970s (Labov, Reference Labov, Fasold and Schiffrin1989). Even more striking is the fact that the traditionally lax set before intervocalic nasals is tense almost half the time. As a whole, this pattern corresponds to the characterization projected at the outset: African American communities approximate the phonology of the surrounding northern city.

Table 1. Percentage of tensing of short-a for 30 African Americans in West Philadelphia

Reprinted, with permission, from Henderson (Reference Henderson1996).

We can now use the locally weighted regression technique to display the phonetic contrast between the traditionally defined tense and lax classes in Philadelphia across the century. Figure 6A displays the contrast by date of birth of the mean values of tense /æh/ versus lax /æ/ for the white mainstream population of the PNC. The two distributions are widely disjunct, although the distance appears to be diminishing in recent decades as the tense class becomes slightly lower. LRF showed that this reduction is primarily the result of younger speakers with higher education switching to the nasal system in which vowels are tensed before all and only nasal consonants. In contrast, the diagram for African Americans in Figure 6B shows that the distinction has practically disappeared as a consequence of the opposite upward movement for the lax class. The broad overlap of the gray 95% confidence areas indicates that there is no significant difference between /æh/ and /æ/.

Figure 6. Locally weighted regression analyses of tense and lax short-a classes in Philadelphia by date of birth for Whites and African Americans. Gray area = 95% confidence areas.

To illustrate the disjunct character of the traditional opposition in the white community, Figure 7 shows the FAVE analysis of tense and lax short-a for PNC subject Jean B., a 60-year-old working-class woman interviewed by Fruehwald in 2006. The downward-pointing triangles represent the traditionally tense class. Words with nasal codas are highlighted, displaying no substantial differences from the unhighlighted tokens before voiceless fricatives. One instance of the tense class—grand—approaches the lax class, a phonetic effect of coarticulation of the initial obstruent-liquid cluster that has been frequently noted (Labov et al., Reference Labov, Ash and Boberg2006; Labov, Reference Labovsubmitted).

Figure 7. Traditional short-a split in the speech of Jean B., age 60 (in 2006).

MEASURES OF BIMODALITY

One useful index of this disjunct character is Ashman's D in (2)—a measure of bimodality expressed by the absolute difference of the means divided by the square root of the average of the two standard deviations (Ashman, Bird, & Zepf, Reference Ashman, Bird and Zepf1994; Gnedin, Reference Gnedin2010).

1. (2)

   $$D\equiv {\vert \mu_1 - \mu_2\vert \over \lsqb \lpar \sigma_1^2 + \sigma_2^2 /2\rsqb ^{1/2}}$$

The value of D for distance along the front diagonal in Figure 7 is 3.439. For a clear separation like that of Jean B., bimodality ranges between 3 and 4.

Figure 8 displays the distribution of this bimodality measure for all PNC speakers with more than 7 tokens of /æh/, including 293 whites and 33 African American speakers. The range is from close to 0.02 to 7.91. It is evident that the black columns are heavily concentrated in the lower end of the distribution. From the diagram, it would appear that the African American speakers have absorbed the short-a system of one particular subtype of the white mainstream pattern, presumably the people they are in closest contact with. But this proves to be an understatement of the differences. The whites with low bimodality are actually those found by LRF to have switched abruptly from the traditional split short-a system to the nasal system, in which all short-a before nasal consonants are fronted and raised along the front diagonal and all others are retained in low front position (Labov et al., Reference Labov, Ash and Boberg2006:174–175). This shift to the nasal system is most characteristic of college students in nationally oriented universities and high school students oriented in that direction. Figure 9 shows the two aspects of this conversion from the traditional Philadelphia system in the speech of a Penn freshman who graduated from an elite high school in Philadelphia. The left-hand figure shows the redistribution of traditionally tense vowels represented by downward triangles. All tokens before front nasals in closed syllables—highlighted here—are retained in upper mid position, with one exception, a token of the word man. All before voiceless fricatives—not highlighted here—are found in low front position. The right-hand figure shows the redistribution of traditionally lax vowels. Those before nasals are tensed: intervocalic (hammer, Danny, Miami) and velar (angry, with one exception).

Figure 8. Distribution of bimodality values of Ashman's D for /æh/ versus /æ/ for PNC speakers with more than 7 tokens of /æh/. Black = African American (n = 33); white = White (n = 293).

Figure 9. The nasal system of Alex P., University of Pennsylvania freshman, 2012.

The same measure of bimodality, Ashman's D, can be used to express the extent to which a speaker has converted to the nasal short-a system. Figure 10 shows bimodality for the traditional system on the horizontal axis and bimodality for the nasal system on the vertical axis. The speakers are divided into seven types. The white majority is indicated by gray diamonds. African Americans are divided into two groups as described earlier. Hatched squares indicate speakers drawn from the LING560 studies of mixed neighborhoods (AA), while hatched triangles designate speakers in the 1986 studies of African American networks in North Philadelphia (AAVE). Vertically striped diamonds indicate the 15 Hispanics who are included here, from LING560 studies of largely mixed neighborhoods. The open symbols show college students from three Philadelphia high schools that lead in the development of the nasal system (all but one are White).

Figure 10. Bimodality measured by Ashman's D for the traditional Philadelphia system (horizontal axis) against bimodality of innovative nasal system (vertical axis) for White mainstream speakers, African Americans, Hispanics, and graduates of three Philadelphia high schools.

One difference between African Americans and Whites in Figure 10 is evident from distribution on the horizontal axis: 203 of 293 Whites have values greater than 2.5, but this is true for only 3 of the 33 African Americans. The 15 Hispanic speakers in the PNC are even more clearly grouped to the left. Table 2 shows the results of a mixed-models regression analysis of these data. Fixed factors ethnicity, date of birth, and sex are significant in that order. In the ethnicity group, Hispanics are even further distanced from the traditional short-a system than African Americans are.

Table 2. Mixed-models regression analysis of bimodality (/æh/ vs. /æ/) for PNC (N = 293)

Note: Significance level in parentheses. Random factor: speaker.

OUTLIERS

A great deal can be learned from an examination of the small number of African American outliers in Figure 10. One is the AA square located at 4.23, 0.55, labeled with the number 1. This is Gloria S., interviewed in 1981 at the age of 60. Her family was among the first African Americans to move into the white neighborhood of South Philadelphia some 50 years earlier.Footnote ⁸ She displays a precise and carefully articulated style of speech, using none of the grammatical features of AAVE.

1. (3) So then maybe you've made a new recipe or you've discovered something in another field that might make you famous for life. You can't tell. But try. You keep on going until life stops in one form or another, because we all know we're going to die one day. But just because you're old or just because you're sick, just because you're handicapped, you're supposed to not keep trying and keep doing? And I think the more you do, the more you're able to do. I really do.

A remarkable opposition is seen in the placement in Figure 10 of two best friends from the integrated Mount Airy neighborhood, interviewed together in their early 20s in 1980. They are located by two other AA squares: Jerome, labeled 2 at 2.59, 1.46 and Burt, labeled 3 at 1.45, 0.97. The two friends disagreed about the extent of integration and everything else on the issues of black/white relations, with Jerome taking the most positive view.

1. (4) Fine, because one reason is, Mount Airy was, you know, always had black and white. I mean, y'know the whites, if they moved, they didn't move too far. I mean it wasn't bad, it was still a good neighborhood, it was a neighborhood that never went deep under, y'know, like poverty or despair or whatever. I don't know. …

Burt's position was almost the opposite.

1. (5) I lived among white people all my life. Don't hate ’em. I don't like some of ’em, but I don't hate ’em. Y'know. It was like uh … color doesn't bother me. But when you walk up in Chestnut Hill, honest to God, if you walk on one side of the street, you get a weird look every time you wanna go… . And people can say that it doesn't bother them, but it's bullshit if they say that. Because just somebody jus’ looking at you weird, like you're a criminal.

The difference in outlook on race relations is reflected in the two friends’ use of Philadelphia phonology, which is immediately evident on listening. Jerome sounds very much like a white Philadelphian; Burt does not. Figure 11 makes this comparison, first showing how Jerome has a clearer separation of tense and lax short-a than Burt does. The lower two diagrams add the mean values for the back upgliding vowels (aw) and (ow). Jerome's system shows a fronted and raised nucleus for (aw), while Burt's (aw) is well back of center. Jerome has strongly fronted (ow), but for Burt, (ow) is a back vowel.

Figure 11. Mean fronting of /aw/ and /ow/. Comparison of Philadelphia vowel systems of Jerome L. and Burt C.

In Figure 10, the AAVE speakers indicated by hatched triangles include some speakers who had extensive interaction with whites—political activists, confidence men, and musicians. Ash and Myhill (Reference Ash, Myhill and Sankoff1986) found that members of the AAVE group with high levels of cross-racial communication had low levels of absence of verbal and possessive /s/, although they were otherwise fully integrated into African American style and culture. One such speaker is the outlier labeled 4 at 1.26, 1.79 in Figure 10, who shows a strong shift toward the nasal system. He is is Steve P., a musician interviewed in 1981 when he was 28 years old. His profession involved him in continual interaction with people of different ethnic and linguistic backgrounds. At the time of the interview, he was about to travel to Germany where he had a number of jobs lined up. At one point in the interview, he underlined the advantages of being a black man surrounded by whites:

1. (6) I figure man I have to leave town to make it. Because, you know, well I been—I been told it's very nice over there for people of my color. You know? Plus, uh, I heard the blacker, the better.

INLIERS

To a some degree, Figure 10 shows the approximation of African Americans to the surrounding white system that was found in the review of the previous literature. The reverse aspect appears in the concentration of minority speakers in the lower left of Figure 10: no whites are to be found in the region limited by 1 on the horizontal axis and .5 on the vertical axis.Footnote ⁹ A prototypical speaker in this region is Jackie C., identified with the black triangle labeled 5 at 0.15,0.44. Her short-a data is shown in Figure 12, displaying an effective merger of tense and lax categories. The phonetic character of that merger involves a general raising of all short-a words but avoids the strongly fronted domain of the white mainstream tensed style, so that the contrast between short /e/ and short /æ/ is small in the F1/F2 dimensions, enhanced by a small difference in mean duration (87 msec vs. 105 msec).

Figure 12. Short-a in the vowel system of Jackie C.

The phonetic pattern found among the speakers in the lower left corner of Figure 10 resembles the pattern reported by Jones in Michigan in Figure 2. The raising of short-a to a nonperipheral, lower mid vowel may be a general characteristic of dialects that reject the peripheral raising of the white community. A similar pattern has been noted among Hispanic speakers in New York City.

Jackie C. is a prototypical speaker of the AAVE community in several respects. The following extract from her speech shows many characteristic features of AAVE, including had as a simple past marker, axe for ask, and negative concord. The relatively lax, merged short-a vowels are found in grandmother, axe, back, and smack.

1. (7) Yeah I remember when I when, one time I took her over to the house and my grandmother was tellin’ me, wa— I shouldn't be bringin’ people over her house. She was arguin’ with Pam then and, when I had brung the baby home, The first day I brung the baby home, and Pam came over, and Pam just went to the, um, to the crib and just looked at the baby.

   And my grandma was fussin’ at her sayin’ “You might carry all kinds of germs,” you know, she was lookin’ at her, bus'in’ on an everything, [Yeah?] and— yeah, you can axe her for herself! She was bustin’ on an’ tellin’ us the way “You sh— ain't got no business bein’ around that baby, that baby just got home! I don't want that baby catchin’ no germs.” She ain't say nuttin’. But then I told Pam, I said “Pam, you know what, you should say something smart right back to her.” Then that's when my aunt smacked me, ’cause my aunt was sittin’ right there. Said, “Don't go back talkin’ back to your grandmother” No, my aunt don't like my grandmother. But yet, she don't like me talkin’ smart to her either.

In addition to the phonology and grammar of AAVE, Jackie displays an immersion in the speech events characteristic of the AAVE community: busting, back talking, and talking smart. At this stage in her life, Jackie C. was prototypical of the core younger speakers of AAVE. In terms of language and in terms of social interaction, she is an inlier as opposed to the outliers Gloria S. and Steve P. Instead of a tendency to approximate the surrounding system that we projected from earlier work, her short-a pattern shows no influence of the white mainstream system that surrounds her North Philadelphia speech community.

The major factor that appears to be operating here is the amount of communication across racial lines (Ash & Myhill, Reference Ash, Myhill and Sankoff1986; Myhill, Reference Myhill1988). The limited sharing of linguistic patterns is a logical consequence of the residential segregation of the city. Children growing up in African American neighborhoods have little personal contact with speakers of the white mainstream community. Table 3 is from Hershberg, Burstein, Ericksen, Greenberg, and Yancey's (1981) study of Philadelphia history, showing the proportion of a person's census tract that consists of the same group. Over a century, segregation of Blacks has steadily increased, while the figures for others have steadily declined. In American Apartheid, Massey and Denton (Reference Massey and Denton1993) point out that America has achieved a higher degree of residential segregation than South Africa ever did. The extent to which blacks live in neighborhoods that are predominantly black jumped from 27.3 in 1930 to 75.6 in 1970 in Philadelphia and similar increases were found in Atlanta, Boston, Detroit, Los Angeles, and San Francisco (Massey & Denton, Reference Massey and Denton1993:Table 3.4). Even in the mixed neighborhoods of Philadelphia like Mount Airy and Port Richmond, most African Americans grow up without any daily interchange with white speakers, and contact is primarily diffusion across adult lines. Given what we know about adult language learning abilities, it is not surprising that the short-a pattern of most African American Philadelphians does not even approximate the complex pattern of the white mainstream community but makes only passing reference to it.

Table 3. Indices of dominance for five ethnic groups in Philadelphia from 1850 to 1970

Reprinted, with permission, from Hershberg et al. (Reference Hershberg, Burstein, Ericksen, Greenberg, Yancey and Hershberg1981:Table 8).

The reduced short-a system common to African American and Hispanic speakers in the lower left corner of Figure 10, with a single /æ/ category, can be viewed in two ways. It may represent the transmission of a prior pattern independent of the traditional Philadelphia system or the result of a gradual diffusion from the Philadelphia system that progressively loses detail. The second possibility aligns the Philadelphia development with the proposed explanation for the origin of the general raising of short-a in the Northern Cities Shift of western New York, as a simplification of the many disparate short-a systems that were brought into contact in the building of the Erie Canal (Labov et al., Reference Labov, Ash and Boberg2006:216). This calls up a somewhat paradoxical generalization: Faithful transmission preserves the irregularities of a system produced by lexical diffusion, but less faithful diffusion can reduce those irregularities to a more regular phonological pattern. This indeed is what seems to have happened in the diffusion of the New York City short-a system to other regions in the 19th century (Dinkin, Reference Dinkin2009; Labov, Reference Labov2007).