Phonetic Convergence in Spontaneous Speech
Principal Investigator: Grzegorz Dogil2,3 & Antje Schweitzer3
Researchers: Daniel Duran3 & Natalie Lewandowski2,3 & Antje Schweitzer2
2 Phase 2
3 Phase 3
Project A4-N investigates phonetic convergence between conversation partners in spontaneous dialogs. Phonetic convergence in general is the process of adapting one's speech to that of an interlocutor, i.e. the perception of the interlocutor's speech affects a speaker's current production targets. Convergence is seen as a dynamic process causing continuous updates in the respective partner's speech. The interlocutor's preceding utterances bias a speaker's productions with respect to those phonetic properties that are underspecified in the given context. Thus, convergence is a prototypical case of incremental specification, and it happens naturally.
In the previuos phase, we investigated convergence by building and analyzing the GECO database of spontaneous dialogs including participants' mutual ratings of likeability and esteem as well as personality measures.
In phase 3, we have shifted our focus and explore the relationship between attentional capacities and convergence. This is motivated by results from N. Lewandowski's (2012) dissertation: She found that the amount of convergence (in a second language context) significantly correlates with the phonetic talent of the participants. She proposed that the underlying mechanism for adaptation highly depends on the speaker’s ability to pay attention to fine phonetic detail. Talented speakers are thus assumed to focus their attentional resources more on the detailed form of the signal. We investigate this hypothesis further. To this end we have developed a 3D computer game for testing speech perception subconsciously. In a science fiction setting, participants have to learn to differentiate between humans and aliens. Instead of just playing the stimuli to participants as in classical perception tests, in this game the utterances of the characters constitute the stimuli. Humans and aliens look alike, but we have modified their utterances so they differ with respect to various phonetic parameters. Participants who can detect these differences easily (who pay more attention to phonetic detail) score higher in the game.
We have extended the GECO database by adding new data including new dialogs and accompanying test results of our subjects from the newly developed attention test, as well as from classical attention and memory tests. The new data include data from female and male participants; there are same-gender and mixed-gender dialogs. The extended database will be published at the end of the project.