The variability in regional distribution and population ecology of
both Europeans and Africans in slavery-era America has long necessitated
an investigation into the issue of regional distinctions in Early African
American English (AAE). Much recent work has speculated on the likely
linguistic outcomes of sociohistoric factors, such as the period of
settlement, African–European population ratios, and type of
agriculture (Mufwene, 1996, 2000; Rickford, 1997; Singler, 1998; Winford,
1997; Wolfram, 1999; Wolfram & Thomas, 2002). In particular,
distinctions have been drawn between the Deep South states of Louisiana,
Mississippi, Alabama, Georgia, and South Carolina, with their larger
plantations and larger African-origin populations, and the remaining
southern states. However, as we have pointed out elsewhere (Van Herk,
2002:14; Walker, 2000:23–24), although sociohistorical evidence
can serve to inform linguistic reconstruction, it can never give us
definitive answers about the linguistic consequences of a particular
situation. To address these issues, we must supplement sociohistorical
information with linguistic evidence.
Unfortunately, such evidence is in short supply. Some work on Early
AAE is based on data too sparse to permit detailed regional distinctions
(e.g., Bailey, Maynor, & Cukor-Avila, 1991;
Montgomery, 1999), some is based on data drawn
from African American diaspora communities, in which the multiple or
incompletely documented input populations complicate regional analysis
(e.g., Poplack & Sankoff, 1987 and Poplack
& Tagliamonte, 2001 for Samaná,
Dominican Republic; Poplack & Tagliamonte, 1991, 2001 for Nova Scotia,
Canada; and Singler, 1989 for Liberia, West
Africa); and some is based on data produced by transcribers or reporters
who were unlikely or unable to note the distinctions relevant for our
purposes (e.g., Dillard, 1972; Schneider, 1989; Viereck,
1989).
This article tests the possibility of regional variation in Early AAE
in a corpus that is both historical (produced during the period of
slavery) and primary (produced by African Americans), as well as
sufficiently large, diverse, and well-documented to allow analysis of its
contents along both linguistic and social (especially regional) axes. The
Ottawa Repository of Early African American Correspondence
(OREAAC) consists of over 400 letters written by ante-bellum African
American settlers in Liberia (Van Herk & Poplack,
2003). The limited literacy of the letter writers leads to the
retention of linguistic variability, paralleling in some cases that
described for spoken AAE (Poplack & Tagliamonte,
2001; Van Herk & Poplack, 2003).
The linguistic variable we examine here, the nonstandard use of verbal
-s in the present tense, is ideally suited to an investigation of
regional diversity in AAE. It has been attributed to a range of origins
associated with specific linguistic conditioning that can be empirically
tested, and it exhibits variability in both historical and recent written
AAE (Funkhouser, 1973; Montgomery, 1999; Montgomery,
Fuller, & DeMarse, 1993). Unlike modern Standard English, in
which only third-person singular verbs are marked, AAE shows variable
marking across the grammatical paradigm, as illustrated in (1).
1 Examples taken from the OREAAC
are identified by the writer's name and the American Colonization
Society microfilm reel, volume, and letter number.
We begin by providing a brief overview of the treatment of regional
variation in Early AAE, before describing the OREAAC and discussing our
analytic methods and the results of our multivariate analyses. The
particular emphasis of this article is on the linguistic conditioning of
verbal -s, comparing writers who originated in the Deep South
with those from other areas.
REGIONAL VARIATION IN EARLY AFRICAN AMERICAN
ENGLISH
Most work on regional variation in Early AAE is based on
sociohistorical reconstruction, largely aimed at determining the
demographic and social conditions that may or may not have been conducive
to the formation and maintenance of a plantation creole (e.g., Mufwene,
1996, 2000; Rickford, 1997; Winford, 1997, 1998; Wolfram & Thomas, 2002). Schneider and his
associates (e.g., Kautzsch & Schneider,
2000; Schneider, 1989) provided a
linguistic focus in their work on the Ex-Slave Narratives (Rawick, 1972, 1979), which are transcribed interviews with
elderly former slaves conducted in the 1930s and 1940s. Based on the
presence or overall rate of use of various forms, this work proposed
regional distinctions delimiting the likely Gullah-speaking area (Kautzsch & Schneider, 2000) and also
differentiating Tennessee from its southern neighbors (Schneider, 1989:251–255). Aside from the general dangers
inherent in the use of secondary data, there are specific problems for
regional differentiation resulting from certain attributes of the Ex-Slave
Narratives. The reliability tests established by Schneider (1989) for the inclusion of transcriptions eliminates
the crucial states of Virginia and Louisiana from consideration, and the
transcriptions are incapable of capturing the relevant phonological
conditioning on verb marking (Schneider, 1989:65–70, 84). In addition, the focus of these
studies on the simple presence or absence of nonstandard forms, or their
overall rates of use, limits their utility in employing regionally
diagnostic constraints on the choice of forms.
Primary data are examined in Singler's (1999) work on verbal -s in diaspora
communities in Liberia, West Africa. His analysis of modern Liberian
Settler English (LSE) showed constraints on -s similar to those
of other diaspora communities (Poplack &
Tagliamonte, 2001), except in Sinoe County, an isolated community
in southern Liberia. Here he found that his fifteen informants had very
low rates of -s-marking, and five had none at all. This led him
to conclude that there were differences in the input varieties to Sinoe
and to the other diaspora communities, including the rest of Liberia.
Because the original Sinoe settlers were largely from the Deep South
states of Mississippi and Louisiana, he argued, the language of Sinoe
represents a distinct variety found only on the plantations of the Deep
South. The language of the Liberian settlers outside of Sinoe, as well as
that of other diaspora communities, would thus be descended from
European-influenced varieties that existed only in marginal agricultural
areas of the early United States that featured a lower ratio of African
American plantation slaves to Europeans. Such an interpretation is
consistent with the “creolist” viewpoint of some contributors
to the debate over the origins of contemporary AAE, who argued that its
salient features derive from a prior plantation language which was very
un-English-like and probably creolized (e.g., Bailey,
1965; Baugh, 1980; Dillard, 1972; Holm, 1984;
Rickford, 1977; Stewart, 1967, 1968, 1973).
An alternative interpretation is that changes have occurred since the
diaspora, perhaps as a result of contact with other language varieties.
This interpretation, consistent with the principles of the comparative
method (Meillet, 1967; Poplack, 2000; Tagliamonte,
2002), would lead us to conclude that it is Sinoe that has changed,
not that every other variety has changed in parallel ways. Singler
dismissed the possibility of contact-induced change (beyond a few
discourse markers and calques) because of hostility between the Sinoe
settlers and their neighbors, which even led to a war in 1857. However,
this argument can equally be made for just about every other diaspora
community, all of which have been socially distant from and often in
conflict with their neighbors (cf., J.S.G. Walker,
1992). Ideally, resolving this issue would involve the use of data
from an earlier time period, so that conflicting hypotheses over the
direction and degree of post-diaspora change could be tested.
DATA: THE OTTAWA REPOSITORY OF EARLY AFRICAN
AMERICAN CORRESPONDENCE
The Ottawa Repository of Early African American
Correspondence (OREAAC) is a collection of 427 letters written by 235
semiliterate African American settlers in Liberia between 1834 and 1866,
and is housed in the Sociolinguistics Laboratory at the University of
Ottawa. As described in detail in Van Herk and Poplack (2003), these letters were selected from among 191,000
documents in the microfilmed archives of the American Colonization
Society, held by the Library of Congress. A rigorous protocol permitted
the selection and verbatim transcription of letters only from those
informants who were the least literate and wrote the least frequently.
These letters feature a range of nonstandard forms that are closely
associated with AAE, including zero copula, r-lessness,
consonant-cluster reduction, vowel lowering, the
pin/pen merger, preverbal bin,
was/were leveling, negative concord, and the absence
of overt marking for plurals and possessives. Perhaps more remarkable is
the finding that not only the forms, but also the linguistic conditioning
of their use, match the results from studies of the spoken language.
Although some AAE features seem to be affected by either the acquisition
of partial literacy or the writing process (notably, negation and most
contractions; cf., Kautzsch, 2002; Montgomery, 1999; Van Herk,
1999), others seem to retain their frequency and conditioning.
Three aspects of the OREAAC make it a particularly suitable corpus for
dealing with these issues.
First, the OREAAC represents a direct (albeit written) view
of the input to the Liberian communities, in that the writers are not just
the descendants of the Liberian settlers, they actually
are the settlers. Most OREAAC correspondents left America between
the ages of 18 and 49, writing soon after their arrival in Liberia. As
such, the language of their letters can be taken to represent their
vernacular, acquired in the United States. This corpus thus provides an
unparalleled opportunity to compare the linguistic input to a diaspora
community with the language spoken there today.
Second, the record keeping associated with the Liberian settlement
gives us detailed demographic information about the majority of the OREAAC
correspondents, gleaned from 19th-century ships' logs and the 1843
Liberian census. As Table 1 illustrates, most
OREAAC correspondents, like most African Americans, were described as
illiterate. Most had been slaves before emigrating to Liberia. Over 60% of
the informants for whom we have information were slaves who had been freed
by their masters, or who had purchased their own freedom. Almost all (90%)
were from the slave states of the South, including large numbers from the
Deep South states, which are sometimes claimed to have produced the most
divergent African American English (i.e., Louisiana, Mississippi, Alabama,
Georgia, and South Carolina). Finally, as far as we can determine by
extrapolating from the slave status of OREAAC writers and the occupations
of slave and free Liberians, most had been agricultural laborers in the
United States. On all these axes, the majority of the OREAAC
correspondents match the demographic characteristics of the majority of
African Americans of the time. Van Herk and Poplack (2003) discussed in detail the status of the OREAAC
correspondents as representative of ante-bellum African American society,
addressing, in particular, issues of literacy and regional origin.2
It is obviously crucial to our arguments that
the attributions of state of origin of these informants be reliable, given
that slaves were often sold from one area into another. Textual evidence
in the letters themselves supports the evidence derived from ships'
logs and the census, in some cases confirming multiple generations of
residence in one state. An exception is the case of informants from free
states such as Illinois, who sometimes mention their (southern) slave
status, but these informants are very few.
Reported demographic characteristics (in percentages) of OREAAC
correspondents, the overall Liberian population, and the African American
population in the United States ca. 1850
Finally, and of greatest importance, the Liberian demographic
information can be applied to a different purpose. Our knowledge of the
state of origin of most OREAAC correspondents allows us to test claims
directly with respect to the language of 19th-century AAE and its regional
diversity. By comparing the linguistic conditioning on variability in
letters written by settlers from the Deep South with those from elsewhere,
we can determine the degree to which their underlying linguistic systems
may differ.
METHOD
From the OREAAC, we extracted those tokens implicated in
-s-marking and coded each token for a series of factors
hypothesized in the literature to condition the variation. The statistical
significance of each factor group and its direction and magnitude of
effect were determined by the multiple regression procedures incorporated
in goldvarb 2.1 (Rand & Sankoff,
1990). By comparing the ranking of factors between writers of
different regional origins, we can determine whether or not they partake
of the same linguistic system.
Defining the variable context
Following recent work (Poplack & Tagliamonte,
2001; Walker, 2000), we defined the
variable context of verbal -s as all verbs with reference to
present time. We excluded any verbs that either clearly had past
reference, and might therefore be subject to competing variable processes
of zero-marking in the past (Van Herk, 2002), as
in (2a), as well as temporally ambiguous forms (2b).
In addition to regular simple present verbs (1a–c), we retained
for analysis irregular verbs (have, be, do)
with simple present tense morphology, including copular (3a), progressive
(3b), and perfect constructions (3c).
This protocol resulted in a total data set of 2,873 simple present
verb tokens, each of which was coded for whether or not it was marked with
-s (or, in the case of irregular verbs, by whether it occurred in
the equivalent form: has, is, does). In
addition, for each token we coded a number of factor groups hypothesized
to affect the variable use of verbal -s in previous studies.
Linguistic factor groups
Phonological context
Studies of Early AAE have reported phonological conditioning of
-s, though the effects are inconsistent across varieties and
grammatical persons (Poplack & Tagliamonte, 1989:64–65, 1991:329,
2001:188–190; Schneider, 1989:70–71; Singler,
1997; Walker, 2000:142). We coded each of
our tokens as to whether the preceding segment was a vowel (4a), a
sibilant (4b), or other consonant (4c), and whether the following segment
was a vowel (4d) or consonant (4e).3
We also
coded the phonological environment for following pause, as indicated by a
full stop (.). However, because it is unclear whether such
“pauses” correspond to spoken pauses, we have excluded such
tokens from consideration.
Type of subject and adjacency
The type of subject is an important consideration in verbal
-s, in terms of both the grammatical status of the subject and
its adjacency to the verb. Full noun phrase (NP) subjects favor
-s over pronominal subjects in most varieties of Early AAE
(Bailey et al., 1989:330; Montgomery &
Fuller, 1996:217–218; Montgomery et al.
1993; Poplack & Tagliamonte, 1989:66, 1991; Walker, 2000:346–347), but not in Liberian Settler
English (Singler, 1997).4
Van Herk (2004) argued that over
time, cumulative phonological tendencies have eroded the NSR effect in
AAE; a study of a contemporary (albeit isolated) variety of AAE by Wolfram
and Thomas (2002) found little evidence of a
subject-type effect among their elderly informants.
The
separation of the verb from its subject also promotes
-s-marking
in Early AAE (
Poplack & Tagliamonte, 2001;
Tagliamonte & Poplack, in press;
Walker, 2000). This constraint has been linked to the
Northern Subject Rule (NSR), which originated with northern British
varieties of English and which requires
-s with plural
present-tense verbs unless they are immediately preceded by a pronominal
subject (Ihalainen,
1992:221–222). Each
token was coded for whether the subject of the verb was an adjacent
pronoun (5a), a nonadjacent pronoun (5b), an adjacent noun or noun phrase
(5c), or a nonadjacent noun or noun phrase (5d).
If Early AAE from the Deep South differs from that of other areas by
being less English-like, we would expect the NSR effect to be absent.
Grammatical person
Studies of Early AAE have generally shown that -s is found
most in third-singular, less in third-plural and least in other
grammatical persons (Montgomery et al., 1993:345; Poplack & Tagliamonte, 1989:62–63, 1991:328;
Schneider, 1983:103, 1989:70; Singler, 1997). The
subject-type effects discussed previously are sensitive to grammatical
person, being most evident in third-person plural contexts. We coded each
token for whether the grammatical person was third-singular, third-plural
or nonthird.
Verbal aspect
The linguistic effect that has received the most attention in the
discussion of verbal -s in AAE is verbal aspect, undoubtedly
because of its relevance to the English-based creoles to which AAE has
often been linked (e.g., Holm, 1988:148–168; Winford,
1993). Several early studies (Bickerton,
1975; Brewer, 1986; Pitts, 1981, 1986) argued that
verbal -s represents the relexification of an earlier preverbal
marker of nonpunctual (habitual or durative) aspect. The habitual effect
has been reported for most varieties of Early AAE (Poplack &
Tagliamonte, 1989:68, 1991:329–330, 2001;
Singler, 1997; Walker, 2000, 2001). Following
recent studies (Poplack & Tagliamonte 2001;
Walker 2000), tokens were coded as habitual (6a)
or nonhabitual (6b) on the basis of the aspectual reading of the sentence
rather than on the nature of the lexical verb.
Verb type
One consideration that has not received much attention in the
literature is the morphological status of verbal -s. If it
represents a relexified earlier aspectual marker, we might expect regular
verbs (which mark person/number agreement solely via suffixation) to
exhibit linguistic conditioning differing from that of irregular verbs
(which additionally mark agreement through vowel change and suppletion).
If -s is simply the phonological instantiation of a more general
process of agreement, constraints between the two types of verb should be
the same.5
One caveat is that, unlike the
situation with respect to past tense marking (cf., Van
Herk & Poplack, 2003), very few verbs in English have irregular
present tense paradigms.
Each token was coded for whether the
verb was irregular (i.e.,
have,
be,
do, as in
(7a) or regular (7b)).
Social factor groups
Reported literacy tests the degree of potential interference
from informants' knowledge of the agreement rules of the standard
written variety. The reported literacy of the informant is reduced here to
low (informants whose literacy was described in source documents
as “unknown,” and who have been presumed in the literature to
be illiterate) or medium (informants whose literacy was described
using such terms as “reads,” “writes,”
“reads and writes,” “spells,” or
“good.”).6
As Van Herk and
Poplack (2003) discussed, the existence of so
many letters written by people assumed by earlier authors to be illiterate
supports recent historical research (Anderson,
1988; Cornelius, 1991) that questions
long-held assumptions that ante-bellum African American literacy was
extremely rare, and that most African Americans learned to read from
whites.
Slave status tests whether rates of -s are higher
among free blacks, who may have had greater access to standardizing
linguistic forces, than among slaves. Here we distinguish only between
freeborn and nonfreeborn writers, the latter including both those who had
purchased their freedom and those who had been manumitted.
The state of origin tests for regional differences, not only
in the overall rate of -s, but also in terms of the hierarchy of
language-internal constraints on -s variation in each region. For
this analysis, we divided informants according to whether their state of
origin is Deep South (Louisiana, Mississippi, Alabama, Georgia, and South
Carolina) or non-Deep South (Kentucky, North Carolina, Oklahoma,
Pennsylvania, Tennessee, Virginia, Washington DC, and New York).7
We initially coded for the area of Liberia from
which the letter was written, divided between Sinoe County and elsewhere,
to test Singler's (1999) hypothesis of
regional differences in Liberia. However, because there was a large degree
of overlap between this factor and that of the state of origin (as we
would expect, given Sinoe's history of settlement), this factor group
was not retained in the analyses.
ANALYSIS AND RESULTS
All of the above factors were considered both individually and
together using goldvarb 2.1 (Rand &
Sankoff, 1990). Table 2 displays the
results of a multivariate analysis of all 2,873 tokens, across all
persons, and their distribution across social factors, including both
regular and irregular verbs. The only factor selected as significant is
the degree of literacy, with more literate correspondents favoring
-s. Separating the analysis according to grammatical person, as
shown in Table 3, reveals that the literacy
effect is significant only in third-person singular contexts, where
-s is required in the standard language.8
The greater use of -s in both standard and
nonstandard contexts by the more literate informants parallels Poplack and
Tagliamonte's (1989) finding for
Samaná English, in which informants with greater exposure to school
and/or church used more -s across all contexts.
In
other words, in this case more literacy results in a greater tendency for
writers to obey Standard English rules of verb concord. However, note that
even the literacy of these writers is not great enough to completely
obviate
-s-marking in other grammatical persons.
Table 3 also reveals that writers from the Deep South
favor
-s in (nonstandard) nonthird-person contexts, whereas
writers from outside the Deep South and freeborn writers favor
-s
in (nonstandard) third-person plural. However, across all social factors,
the hierarchy of grammatical person (i.e., third-singular >
third-plural > nonthird) is maintained, as demonstrated by the
percentages of rates of use.
Social factors contributing to the occurrence of -s-marking
in the OREAAC
Social factors contributing to the occurrence of -s-marking
in the Liberian letters, by grammatical person
At first glance, the distinction between Deep South and non-Deep South
writers lends credence to Singler's (1999)
claim that differences in modern-day LSE may reflect different linguistic
systems associated with the different origins of his Liberian Settler
informants. However, the important consideration in comparative analysis
is not the overall rate of occurrence across social groups, which can
fluctuate for a number of (extralinguistic) reasons, but rather whether
different groups (here, regions) share linguistic conditioning, as
evidenced by the ordering and significance of linguistic factors
contributing to the occurrence of -s.
Therefore, the next step in the analysis is to consider whether the
linguistic conditioning is parallel across the two groups of writers. In
addition, the inclusion of both regular and irregular verbs may confound a
purely phonological process (insertion and/or deletion of -s)
with morphological or lexical considerations (suffixation, stem
alteration, suppletion). Therefore, we performed separate multivariate
analyses of -s-marking according to the type of verb, as well as
the writer's state of origin. Table 4 shows
the results of these analyses.
Factors contributing to the occurrence of -s-marking in the
Liberian letters, by verb type and state of origin
First, as noted previously, the OREAAC correspondents always adhere to
the person-number hierarchy observed elsewhere, whereby rates of
-s are highest in third-person singular, lower in third-plural,
and lowest of all in other persons. In Table 4,
this factor is not only selected as significant, but its strength
(instantiated by the range) is considerably greater than the effect of
other linguistic factor groups. This effect holds for both regular and
irregular verbs, and for writers from the Deep South or elsewhere.
Although the effects of subject type/adjacency and aspect are similar
across categories in the percentages, these factor groups have not been
selected as significant for all combinations of verb type and state of
origin. However, recall that the effect of subject type/adjacency is
claimed to hold only in third-person plural. Therefore, it is necessary to
examine the conditioning of -s in each person-number context
separately. Table 5 shows the results of a
multivariate analysis for regular verbs, considering grammatical persons
separately.
Factors contributing to the occurrence of -s-marking in the
Liberian letters, by grammatical person and state of origin (regular verbs
only)
The subject type-adjacency factor group (which tests the Northern
Subject Rule) exerts a significant effect only with third-person plural
subjects, the very context in which it has been found to occur in previous
studies of Early AAE (and other varieties of nonstandard English). Writers
from both Deep South and non-Deep South states avoid -s-marking
in third-plural contexts when the subject is a pronoun adjacent to the
verb, and prefer -s with all other subjects. In fact, the NSR
effect is so strong in both groups that no other linguistic factor group
is selected as significant in third-plural contexts.
In other grammatical persons, the absence of the NSR effect (and,
arguably, the context which permits its application) allows other
conditioning to achieve significance. In particular, preceding
(nonsibilant) consonants exert a strong favoring effect on -s in
nonthird-person contexts. This is the strongest overall effect (with a
range of 36) in the Deep South, and it is categorical in the non-Deep
South. This effect may be accidental or epiphenomenal, because
phonological effects have been inconsistent or insignificant in earlier
studies of -s-marking. Another possible explanation for this
effect is hypercorrection, which we might expect to play a role in written
documents. The preceding segment effect is significant only in
nonthird-person contexts, where -s is infrequent (and forbidden
in the standard) and occurs largely in phonological contexts where
speakers wishing to avoid consonant clusters might be most expected to
reinsert, paralleling an early finding for contemporary AAE (Labov, 1972:221).
The other strong effect in nonthird-person contexts is the favoring
effect of habituals, described elsewhere for other Early AAE varieties.
That this effect achieves significance only in the Deep South might result
from the slightly larger amount of data for that region (especially given
the overall infrequency of habituals in nonthird-person to begin with).
However, note that the aspectual effect also holds in the Deep South for
third-singular contexts, which suggests that this effect is valid.
The linguistic conditioning of irregular verbs has received less
scholarly attention than that of regular verbs, and studies usually
consider the two categories together. Table 6
illustrates the results of a separate variable rule analysis of 1,443
irregular-verb tokens (largely have, be, and
do) in the OREAAC.9
Since
phonological factors are irrelevant in this context, they were not
considered in the analysis.
Factors contributing to the occurrence of -s-marking in the
Liberian letters, by grammatical person and state of origin (irregular
verbs only)
First, note that the results are identical for the Deep South and
non-Deep South writers. The NSR exerts a strong effect in third-person
plural contexts, across region of origin. As we would expect, the NSR
effect does not operate in third-singular, but it does appear to surface
in nonthird-person contexts. Thus, the influence of the NSR effect with
irregular as well as regular verbs suggests that verbal -s is a
form of agreement, rather than a simple operation of suffixation.
The aspectual findings are also identical across regions. Only in
third-singular is aspect significant, but in a direction opposite
to that described elsewhere, with habituals disfavoring. However, we argue
that this apparent effect results from the lexical composition of the
irregular category. Unlike the situation with past marking, which features
an array of irregular verbs (Van Herk, 2002;
Van Herk & Poplack, 2003), the irregular
category for present tense consists entirely of the verbs have,
be, and do. The most frequent verbs, have and
be, which occur largely in nonhabitual contexts, tend to show
standard agreement patterns (cf., Patrick, 1991;
Walker, 2000), leading to a contrary habitual
effect only in third-singular, the only grammatical person requiring
-s in Standard English. Such lexical effects, especially in
irregular verbs, have been shown to underlie a range of apparent aspectual
tendencies, especially given that lexical identity is often inextricable
from aspect (Poplack & Tagliamonte, 2001;
Tagliamonte, 1991; Van Herk
& Poplack, 2003; Walker, 2000).
DISCUSSION
Our findings strongly suggest a number of conclusions. First, the
Northern Subject Rule was clearly operative in the language of African
Americans during the era of slavery, regardless of where in the United
States they came from. Thus, if there were regional distinctions in Early
AAE, they did not include differential application of the NSR, which has
been found in other studies to contribute to Early AAE
-s-marking. Because the OREAAC represents (a written version of)
the input language varieties to the Liberian diaspora communities, we
suggest that contemporary Liberian Settler English should show traces of
the NSR effect, unless later developments have eroded either the
conditioning of -s-marking or its overall rates of use. In the
case of Sinoe, the decrease in -s-marking may be due to a number
of reasons, such as contact with speakers of Liberian Pidgin English or a
second-language variety of recaptured slaves who were integrated into the
settler community. The strong tendency towards consonant-vowel (CV)
syllable structure in the Liberian Settler English of Sinoe (Singler, 1991) may have led to word-final
consonant-cluster deletion, which would have had the effect of eliminating
verbal -s.
In fact, a close examination of the data provided in Singler (1999) shows that the Sinoe speakers who do
use verbal -s show linguistic conditioning remarkably similar to
what we have described for the Deep South correspondents. Although Singler
did not investigate adjacency, he did test for the type of subject. As
Table 7 shows, the “Sinoe Nine” show
a strong subject-type effect, as well as an aspectual effect, parallel to
our findings. Such parallels show that, despite the differing rates of
use, the linguistic conditioning remains constant. Thus, the system
generating this variability is revealed by the linguistic factors
conditioning the use of the form, not simply by its overall rate of
occurrence.
Linguistic factors contributing to the occurrence of verbal
-s in “the Sinoe Nine” (adapted from Singler 1999:5–6)
These findings remind us of an operative assumption of the comparative
method, as applied to variable linguistic evidence: namely, that
similarities across varieties tell us more than differences (Poplack, 2000; Poplack &
Tagliamonte, 2001; Tagliamonte, 2002).
Specifically, shared retention of variable conditioning that cannot be
attributed to universal considerations constitutes evidence of a shared
origin. In this case, similar linguistic conditioning across
widely-separated diaspora corpora is matched in the letters of writers
from across the American South. As in traditional historical linguistics,
we cannot interpret the absence of shared conditioning in widely separated
or differently constituted varieties as proof that the varieties were
never the same, although we rarely have access to the kind of diachronic
confirmation the OREAAC offers.
Other similarities that we note here include the cross-regional
tendency to use -s in nonthird-person contexts after consonants,
which may be evidence of hypercorrection, and to disfavor -s with
habitual irregular verbs, which appears to be an epiphenomenon of lexical
considerations. Deep South and non-Deep South writers also behave
identically in their application of the NSR effect in irregular verbs,
which may provide theoretical insight into issues of how agreement rules
operate in nonstandard varieties of English (cf., Börjars & Chapman, 1998; Henry, 1995).
The one robust regional distinction is the tendency of habituals to
favor -s outside of third-plural contexts, statistically
significant in the Deep South and potentially derived from one or more
aspect-prominent donor varieties. It may be that the habitual effect
achieves maximum strength only in those areas where (almost) all potential
donor varieties contribute such a habitual distinction. That this tendency
is also found across African American diaspora communities testifies to
the speech-like nature of the OREAAC, and further suggests that this Deep
South component is present in all the diaspora communities studied,
apparently even those for which precise demographic information is
unavailable.
The findings described here reinforce many of the earlier findings
with respect to Early AAE -s-marking derived from work on
diaspora communities. However, in this particular case, the application of
variationist methodology to the analysis of a frequent and highly variable
form in a regionally widespread, authentic 19th-century corpus of written
Early AAE has allowed us to investigate regional diversity to a degree not
previously possible. These results suggest that Early AAE, like any
geographically widespread linguistic variety, displays the results of two
opposing but complementary forces. There are subtle regional distinctions
when specific settlement and population ecologies encourage them, but more
important—and impressive—is the overall identity of forms and
conditioning factors determined by shared history and circumstances of
language contact and development.
