The paper investigates how non‑linguistic vocal cues—specifically activity, engagement, stress, and mirroring—can be automatically measured and used to predict outcomes in socially critical interactions. Four signal measures are defined using hidden Markov models and prosodic statistics, and a real‑time implementation on a PDA is described. Three empirical studies are reported: (1) a salary‑negotiation experiment where thin‑slice (first five minutes) measures predict final negotiation scores (overall r=0.57; engagement+stress especially predictive for high‑status participants, mirroring for low‑status participants); (2) a speed‑dating study where the same measures predict romantic, friendship, and business attraction decisions (≈70 % accuracy, with female activity and engagement most informative); (3) an 11‑day campus‑wide audio study linking the engagement measure to betweenness centrality in the social network (r≈0.90). The results show that vocal social signals are distinct, stable, and strongly related to both objective outcomes (e.g., deal size, match formation) and subjective perceptions (impression, cooperation). The discussion argues that these signals constitute a form of “vocal body language” that mediates status, interest, and cooperation, and suggests future wearable feedback devices to make people aware of their own signaling.
Topics
Non‑linguistic Vocal Signals as Predictors of Social Outcomes
Introduces the premise that tone of voice and prosodic style, though often unconscious, can forecast behavioral results across domains such as negotiation, dating, and network influence.
Four Automated Vocal Measures: Activity, Engagement, Stress, Mirroring
Defines each signal, explains computational pipelines (HMM segmentation, turn‑taking influence modeling, prosodic variance, short‑utterance exchange frequency) and how they are z‑scored.
Real‑time Signal Extraction on Mobile Devices
Describes implementation of the measures on a PDA with a one‑minute lagging window, enabling minute‑by‑minute analysis of conversations.
Negotiation Experiment: Thin‑Slice Prediction of Deal Success
Details the salary‑negotiation study (46 gender‑matched dyads), methodology (first five minutes), and key findings (overall r=0.57; engagement+stress predict high‑status outcomes, mirroring predicts low‑status outcomes).
Speed‑Dating Experiment: Vocal Signals and Romantic/Professional Attraction
Presents the speed‑dating study (57 sessions), regression results (≈70 % accuracy), gender‑specific effects (female activity predicts attraction, engagement predicts friendship/business), and the stronger influence of female signals.
Campus Social‑Network Study: Engagement as a Connector Indicator
Summarizes the 11‑day audio collection from 23 subjects, showing that individual turn‑taking patterns are stable and that engagement correlates (r≈0.90) with betweenness centrality, linking vocal signaling to network position.
Discussion: Vocal Body Language, Status Negotiation, and Future Wearables
Interprets the results as evidence that vocal social signals constitute a distinct communication channel influencing status, cooperation, and information flow, and proposes wearable meters for real‑time feedback.