Knowledge and Processing of Morphosyntactic Variation in African American Language and Mainstream American English

Thumbnail Image


Publication or External Link





As people from different social groups come into contact, they must accommodate differences in morphosyntax (e.g., He seem nice vs. He seems nice) in order to successfully represent and comprehend their interlocutor’s speech. Listeners usually have high comprehension across such differences, but little is known about the mechanisms behind morphosyntactic accommodation. In this dissertation, I asked what listeners know about variation in morphosyntax and how they deploy this knowledge in real-time language processing. As a test case, I focused on regularized subject-verb agreement (e.g., He seem nice, They was happy)—which is common in African American Language (AAL), but not in Mainstream American English (MAE)—and compared how listeners adjust their linguistic expectations depending on what language varieties both they and their interlocutors speak.

In Experiment 1, I showed that participants who primarily speak MAE 1) recognize that some speakers use regularized subject-verb agreement, 2) evaluate that regularized subject-verb agreement is associated with AAL, and 3) predict that the subject-verb agreement rules of AAL allow for some patterns (They was happy) but not others (*He were happy). This was accomplished using a novel sentence rating task, where participants heard audio examples of a given language variety, then rated written sentences for how likely a speaker of that variety would be to say them. In Experiment 2, I showed that a similar pool of participants did not merely recognize regularized subject-verb agreement; their knowledge of variation lead them to predict that AAL speakers use regularized forms in an acoustically ambiguous context. Participants heard sentences like He sit(s) still, where it is unclear whether the verb includes a verbal -s due to a segmentation ambiguity. They were more likely to transcribe a regularized form (He sit still) when it was spoken by an AAL-speaking voice than when it was spoken by an MAE-speaking voice. Together, these results indicate that listeners have rich mental models of their interlocutors that extend beyond a general awareness of linguistic difference.

In Experiment 3, I compared bidialectal speakers of AAL and MAE and monodialectal speakers of MAE. On the rating task from Experiment 1, bidialectal participants showed a greater degree of differentiation between sentences that are grammatical in AAL and sentences that are ungrammatical in AAL, compared to monodialectal participants. However, both groups of participants indicated that ungrammatical sentences are broadly more likely in AAL than MAE, contrary to usage patterns in the world. On the transcription task from Experiment 2, bidialectal participants were overall more likely to transcribe regularized subject-verb agreement, but they differentiated between AAL- and MAE-speaking voices to the same degree as monodialectal participants. Both groups were more likely to use MAE subject-verb agreement (He sits still) than regularized subject-verb agreement (He sit still). These results suggest that bidialectal listeners broadly expect regularized subject-verb agreement to a greater degree than do monodialectal listeners, rather than making stronger predictions about a given speaker. Moreover, while bidialectal listeners have a more granular sense of AAL’s grammatical rules, all listeners still favor MAE, likely reflecting MAE’s dominant status.

In Experiment 4, I asked how listeners use their knowledge of variation in subject-verb agreement to guide real-time interpretation of sentences, again comparing bidialectal and monodialectal participants. Participants heard sentences like The duck(s) swim in the pond, where they must rely on the agreement morphology of the verb to determine whether the subject of the sentence is singular or plural, since a segmentation ambiguity makes it unclear whether the noun ends in -s. In MAE, only a plural interpretation is available, while in AAL, a singular interpretation is also available. Participants’ eye-movements were tracked as they looked at and selected images on a screen. Participants were more likely to look at and select a singular image if the sentence was presented in an AAL-speaking voice, compared to an MAE-speaking voice, and bidialectal participants were more likely to look at and select a singular image, compared to monodialectal participants. As with the transcription task in Experiment 3, this suggests that bidialectal participants are broadly more likely to consider the possibility that a speaker uses regularized SVA, compared to monodialectal participants, but their linguistic expectations are not more strongly differentiated based on the grammar of their interlocutor.

These results make it clear that listeners have mental models of morphosyntactic variation, which can be characterized along a variety of dimensions, including the syntax, semantics, and indexicality (social meaning) of a given variable. This can serve as a foundation for future inquiry into the details of these models and the real-time switching and control dynamics as listeners adjust to different varieties in their environment.