Identity and Speech

This series of papers seemed to be designed to teach us that identity is thoroughly involved in the production and perception of speech, contrary to what the study we discussed in class might suggest.

The paper by Podesva et al. examined some of the phonological features of Condoleezza Rice’s speech as it relates to her belonging to several groups with distinctive phonetic characteristics, with the aim of gaining insight into how speech shapes and is shaped by a speaker’s identity. The authors analyze the use of both vowel and consonant sounds and conclude that Rice’s speech is “largely aregional” with some features of African American English, which is consistent with a professional, well-educated African American speaker.

For me, the most notable part of the paper was this sentence in the conclusion: “Her engagement with the standard language market is in some ways a symbolic embodiment of a particularly educated construction of the American Dream.” The claim that Rice’s speech can tell us something about the American Dream demands explanation, but the authors omit any. What vision is Rice’s speech supposed to symbolize? It surprised me to find such a politically charged idea popping up in the conclusion of the paper.

The paper by King and Sumner presents experimental data supporting the hypothesis that words are interpreted in speaker-specific ways. The first experiment, a word association task, suggests that speaker phonetics effects semantic associations, and the second experiment, a semantic priming task, demonstrated that for one speaker, association strength correlated negatively with reaction times. The researchers conclude that listeners incorporate social knowledge about the speaker in processing semantic meaning.

I would be interested to learn more about the statistical methodology of the paper. In particular, what are the “gradient models” in experiment 2, and what methods were used to support the conclusion in the results section?

In Sumner’s article in Science and Society, she discusses the early presence of social representations in acoustic processing. The gist is that extremely early in the process of listening to a speaker, we are already using phonetic clues to infer the speaker’s social identity, and our social representations of speakers’ identities, including biases, influence how we process speech at a low level.

I appreciated the writing style of this article. I think the author made some nice choices about how to present the argument. For instance, the example of the two women listening to a theoretical physicist giving a presentation exemplifies a situation where phonetics, visuals, and expectations were all pertinent, but it also gives readers a sense of why the subject of the paper matters.

In the paper by Sumner and Kataoka, the authors present two experiments targeting the effect of phonetic variation on word recognition. They conclude that frequency of exposure to a word is not the best way to account for the data and that the context of the speech (social weighting) or attention-modulating effects of identity information may prove more fruitful explanations.

This paper relied more heavily on jargon than the rest, and it was fairly difficult for me to understand. Particularly confusing were the central concepts of “clusters” and “encoding,” which were used in unfamiliar ways.

The suggestion that difficulty recognizing the words of speakers with unfamiliar accents could have to do with lack of attention intrigued me, because it goes against my intuition about what is challenging about listening to different accents; when I encounter an accent I find difficult, I always think I am trying my best to understand the speaker and focusing attentively on the sounds I am hearing. Notwithstanding, we have seen in class that humans are bad at noticing linguistic subtleties, including the differences between phoneme variants, so I cannot say for certain that I am paying full attention.