• No results found

3.6 Additional results: extra-grammatical effects

4.2.5 Summary

Theoretical explanations for the place of verbal -s in the AAVE grammar are based on constraints, or lack thereof, that have been found in verbal -s variation in modern AAVE and related varieties. The primary analyses of verbal -s can be generalized to two broad ideas:1) verbal -s is not part of the underlying grammar, due to the lack of phonological and grammatical constraints on its variation (Labov et al., 1968, Fasold, 1972); or2) verbal -s is part of the grammar as an aspectual or agreement marker (indicating present tense, habituality, or narrative present), whose roots can be traced to either Anglican or Creole origins. These are all in the context of verbal -s also being subject to intra- and

inter-speaker variation, resulting in the need for controlling for both context and speaker in any dataset.

4.3 Introduction to results

4.3.1 Overview

In general, internal grammatical constraints do not appear to be significant for the third person singular -s absence at this point, however, they remain a challenge open to future studies.-Alim, 2004

Verbal -s in AAVE is both controversial theoretically and difficult to predict quantita- tively. This dissertation will demonstrate additional constraints on verbal -s, and corrob- orate previous findings that verbal -s is subject to social variation such as style shifting.

In light of these new constraints, many of which are parallel to constraints on AAVE copula, I argue analyze verbal -s is part of the underlying grammar.

4.3.2 Methodology

The data are a subset of the longitudinal Frank Porter Graham Corpus of African Americans from age 1 to 18 in Chapel Hill, NC. The data was coded for verbal -s presence as part of the FPG project in collaboration with NCState, which has been used in studies such as Van Hofwegen and Wolfram, 2010, Renn, 2007, and Renn, 2010. A subset was coded specifically for this project to include subject, subject animacy, and verb; auxiliaries were excluded. Quotations were excluded to avoid any imitated or mock usage of variants, and

to avoid overt codeswitching1. 250 total speakers are included, with 1716 tokens (48% null

-s).

I performed a coding reliability check on 60 tokens, with few repeat speakers (57 sep- arate speakers), and from 50 separate paired interviews. Of these 60 tokens, I coded 55 of them in the same way (31 null form, 24 overt form). The 5 that I did not code the same way as the original FPG coding were all originally coded as verbal -s presence, but I coded them as absence. If my coding was in fact accurate, this type of coding error is expected when the coders are MAE rather than AAVE speakers, because there may a tendency toward coding what you expect in your own grammar (Labov, p.c.).

Like in the MAE contraction and AAVE copula chapters, only tokens whose subjects could reliably be coded by animacy are included. That means that references to animals are excluded, among other ambiguous contexts. There are several phonological contexts that make coding verbal -s presence either impossible or highly unreliable, and are excluded. This includes verbs before words starting with -s or other sibilants, as in 29. Verbs ending in the consonant clusters [st], such asinsistandconsist, are also excluded (30).

Statistical methods are the same as those used in previous chapters. In the mixed mod- eling for verbal -s, I use random effects for the speaker, the verb, and the subject head.

(29) He break-? stuff.

(30) (Like) she be wrong but still she insist-? that she right.

I limit the discussion of verbal -s to main verbs only, as previous research has indicated that auxiliaries act differently than main verbs (Rickford and McNair-Knox, 1994). Main verbhaveis excluded as well, for ease of coding. In various subsections, I subset or exclude certain types of tokens to investigate specific questions as conservatively as possible. For 1Quotations were particularly excluded because some of them are quotes from white teachers, and would

instance, in the transitivity subsection, I exclude tokens that were ambiguous with regard to transitivity.

4.4 The animacy effect

In this section, I look at the core effect in this dissertation: that of animacy on verbal -s. First I demonstrate thatanimacy is a significant and robust quantitative predictor of verbal -s variation. Then, I investigate several other factors to determine if there is evidence for animacy effects being related to other possible explanations, and find that it is a robust and independent effect.

In AAVE, stigmatized features are subject to style shifting, codeswitching, and differ- entiation by sex (among many other effects, such as self-construction of identity). I test if there is any evidence that animacy is correlated with or potentially caused by overt stylistic and identity-based effects like formality of context and sex. The prediction is that a gram- matical or processing effect should not interact with these stylistic effects, but instead be apparent across-the-board. Indeed,animacy effects are evident regardless of style or sex, and do not interact with either, indicating thatanimacy is not part of the social domain of linguistic variation.

We can see that the coefficient of animacy (-0.87) is not as high as that of formality (2.49), but is still quite large.

Past studies have quantitatively focused on social effects on verbal -s variation, which is shown in Table 4.1 of the data used here. Social effects on language in the FPG corpus are discussed more at length in Renn, 2007, Renn, 2010, and Van Hofwegen and Wolfram, 2010. I demonstrate that animacy of the subject head (Table 4.2) is highly significant in determining verbal -s presence or absence, and significantly improves the model (as shown in the LLRT Table 4.3).

Estimate Std. Error z value Pr(>|z|)

(Intercept) -3.48 0.69 -5.06 0.00

Age 0.21 0.04 4.74 0.00

Formality: formal 2.66 0.28 9.65 0.00

Sex: M -0.36 0.26 -1.39 0.16

Table 4.1: Verbal -s selection by social and stylistic factors (N=1755) Estimate Std. Error z value Pr(>|z|)

(Intercept) -2.59 0.70 -3.69 0.00

Animacy: human -0.87 0.21 -4.21 0.00

Age 0.20 0.04 4.58 0.00

Formality: formal 2.49 0.27 9.13 0.00

Sex: M -0.37 0.25 -1.47 0.14

Table 4.2: Verbal -s selection by animacy of subject, as well as social and stylistic factors (N=1755)

Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq)

model0 7 1789.78 1827.33 -887.89 1775.78

model1 8 1780.96 1823.87 -882.48 1764.96 10.82 1 0.00

Table 4.3: LLRT where model1 includes animacy as a factor and model0 does not

4.5 Additional results: grammatical effects