One of the possible explanations for differences across subject length effects is that there is some notion of “default” in the grammar, and in morphosyntactic variation produced by the grammar. This assumes that every variable has an arbitrary default that takes less cognitive processing to produce than the other possible variants. I investigate what the defaults could be for each case study using subject length effects, and for AAVE verbal -s in particular using persistence effects.
Subject length effects
One possible interpretation of subject length effects is that longer subjects produce higher cognitive load, and that under the context of higher load, the grammatical default for a variable is more likely to occur. Under this assumption, we can use results from subject length effects across case studies to point to possible defaults in each variable. In MAE contraction, the longer the subject, the more likely the full form. Thus, the interpretation from the assumptions above is that the full form is the default for MAE auxiliaries. This contrasts with AAVE copula2, where the longer the subject, the less likely the full form
is. The interpretation is that the reduced forms, contracted and deleted, are the default for AAVE copula. AAVE verbal -s results are parallel to AAVE copula, such that longer sub- jects prefer the null variant. These results may point to a critical contrast in the underlying grammars of AAVE and MAE, such that the default variants are different, even in similar variables conditioned by parallel constraints.
Persistence
Another way to analyze possible default status is to investigate the effects of persistence of each variant. Persistence effects3may be a useful diagnostic of variants that are underlying
in the grammar (Tamminga, 2014). I demonstrate that verbal -s is persistent, and present this as evidence that both the overt and null variants are stored in the grammar as vari- ants of verbal -s. Secondly, under the assumption that variables have default forms, I use persistence as a possible diagnostic for identifying the default variant.
Many linguistic variables are conditioned by persistence, or the tendency to reuse a recently-used variant. The data here is from 583 interviews with African American children 2It would be interesting to investigate if there is similar the patterning ofwillandhasin AAVE, which,
like copula, can be full, contracted, or deleted.
3The analysis of persistence in this dissertation represents joint work with Meredith Tamminga, some of
in the Frank Porter Graham Corpus. To measure persistence, each token (target) is coded for the presence or absence -s in the previous token (prime), and the distance between target and prime in orthographic words. Persistence of primes and targets was coded by tagging every token of verbal -s presence/absence with the value of the previous token of verbal -s. Previous tokens were not coded across interruption by interlocuter.
Persistence strength is calculated as the following: for any given subset, relative change with respect to the unprimed variant is the equation in Figure 5.1.
Figure 5.1: Equation for persistence strength
Figure 5.2 shows that third singular -s (N=1773) is subject to a robust persistence effect. The upper line shows that previous absence of -s depresses the probability of using -s subsequently, while the lower line shows that the previous presence of -s increases the probability of subsequent -s. Persistence decays over distance from the prime, as shown in the x axis, which displays the number of orthographic words between the prime and the target. These results indicate that verbal -s is subject to persistence effects, and, under the assumptions previously described, indicates that both variants are encoded in the grammar. These results may also point toward a default variant. Note that at approximately the 50 word lag mark on the x axis, the line representing the null prime plateaus, while the line for the overt prime continues to dive until it converges with the null prime. Rather than converging at the average baseline (the dotted line), null -s appears to be the preferred form. I propose that this supports the default status of the null variant, particularly when combined with the results of the subject length effect.
Figure 5.2: Persistence of 3sg -s by lag distance
The convergence of subject length and persistence results toward the null variant of verbal -s being the default is also compatible with other studies. For instance, it predicts results such as Terry et al.’s, in which higher rates of overt -s are cognitively taxing for some AAVE speakers. However, this would also predict that overt copula would also be taxing in similar ways, which Terry et al. tested for but did not find. More specific research is necessary on the interaction between working memory load, subject length, and possible default variants. The possible existence of default variants that are arbitrary and language specific indicates that attempts to universally predict effects on variation still need to appeal to a basic notion of language-specific grammars.