An interdisciplinary team from the USC Viterbi School of Engineering and the University of Utah pooled their expertise to develop the algorithm, which breaks down voice recordings using established speech-processing techniques. The results include distinct vocal features that analyze pitch, intensity, and acoustic “warbles” that point to emotional reactions. This system of direct vocal analysis also beat out more traditional systems of prediction. Human relationship experts coded the same recordings for conversation qualities they considered either positive or negative like blame and acceptance. Overall, human analysis was less accurate than the algorithm in predicting relationship success.
The algorithm also features an added analysis element since it analyzes the impact of each partner’s tone of voice over time. For more than one hundred couples included in the study, the computer program studied the development of conversations between the two partners. The study confirmed that each partner’s impact on the other was predictable based on tone of voice alone. Analyzing these dual partner dynamics boosted the accuracy of the algorithm’s predictions around probable relationship trajectories.
“What you say is not the only thing that matters; it’s very important how you say it. Our study confirms that it holds for a couple’s relationship as well,” said Md Nasir, a doctoral student contributing to the project. General theories of communication agree that tone of voice accounts for about 38 percent of communication, while the words themselves account for only 7 percent. These findings are something of a cautionary tale, but researchers believe they could create more useful objective behavior measurements for couples therapists. So next time you’re working on your communication skills, you’d be smart to pay attention to the tone of your voice.
