Can your voice reveal how stressed you are? This study shows why short relaxation sessions may not be enough to produce measurable changes in speech-based stress signals.
Study: Assessing effects of vibroacoustic stimulation compared to a guided mindfulness meditation using the biosignal of human speech. Image credit: Tero Vesalainen/Shutterstock.com
Managing stress is vital in today’s world, making stress relief interventions an urgent area of research. Speech prosody, the melody and rhythm of speech, may offer a non-invasive way to detect stress. A study in Frontiers in Network Physiology evaluated the effectiveness of this signal in detecting stress-related changes following relaxation interventions.
Speech patterns emerge as non-invasive stress indicators
High levels of stress disrupt well-being at all levels and reduce economic stability. Stress is commonly linked to depression, which affects 5 % of people globally. Stress also compromises motivation and creativity, reducing productivity. Currently, the stress response is understood as a result of physiological network interactions. This has prompted the development of novel tools for stress recognition.
The current study examines speech prosody as a non-invasive marker of changes in the autonomic nervous system's stress response. Speech prosody readily responds to changes in physiological and psychological network states. Moreover, it is easily and conveniently acquired, allowing relatively unbiased data collection.
Speech prosody changes include alterations in speech frequency, pitch, and intonation occurring with emotional or physical strain and cognitive load. These often accompany other changes, such as increased heart rate, and reflect physiological arousal. The rhythm of speech and the quality of the voice also change with stress. All these changes display distinct profiles in mental versus physical stress.
Stress-related network changes induce sympathetic activation, resulting in stiffness of the speech musculature. This includes the muscles of the larynx (voice box), the respiratory muscles, and those of the tongue, lips, and jaws needed for clear articulation.
Such changes are picked up by artificial intelligence (AI), which can detect stress at accuracies typically reported in the 70–90 % range depending on context and methodology. When coupled with other physiological signals, accuracy remains higher. In earlier research, the authors demonstrated approximately 86 % accuracy in predicting pre- and post-treatment readings using acoustic-prosodic parameters.
Previous work has established the effectiveness of vibroacoustic stimulation as a relaxation technique. This uses audible sound combined with synchronized tactile vibration. The authors sought to determine whether a shorter exposure period to vibroacoustic stimulation would induce measurable changes in speech prosody. They compared this with the effects of meditation, a process of self-reflection without judgment that induces calmness and relaxation. It achieves this in part by increasing parasympathetic activity.
Pilot study compares mindfulness, vibroacoustic, and control groups
The researchers conducted a pilot study with 30 participants, forming three treatment groups: guided mindfulness, vibroacoustic stimulation, and a control group.
All participants read a given passage aloud, first before the treatment, and secondly after one 20-minute session. The readings were then analyzed using multiple parameters to assess acoustic-prosody.
This refers to the acoustic features of speech in combination, which determine the speech sound rather than the content. The aim was to evaluate how the relaxation intervention affected prosodic parameters in the final reading opposed to the initial one, on an individual basis.
Voice quality shifts detected but broader effects inconsistent
The results showed that before/after differences hinged on three dimensions: intonation, vocal effort or voice quality, and loudness.
In terms of intonation, the control group participants read at a higher pitch before the treatment period and at a lower pitch after it, with the post-treatment pitch spanning a broader range. In the other groups, there was no significant difference between the before and after.
The same trend was observed with vocal effort. Control participants' voices sounded more effortful or tense after the intervention period than before. The other two groups did not show this change. In contrast, both intervention groups had a breathier voice post-treatment, which is typically associated with a more relaxed physiological state but, on its own, does not confirm stress reduction.
Loudness decreased after the intervention period in the controls, but not in the other two groups. In the control group, participant voices during the post-treatment reading were 4-5 dB softer than before treatment. The other two groups remained louder after the treatment and showed greater loudness variability than the controls. However, for all groups, loudness variability declined over the course of the experiment.
Overall, the experiment produced mixed, largely inconsistent results, with changes observed across all groups, and the intervention groups largely maintained their baseline speech patterns while the control group showed stronger shifts. The controls spoke in deeper voices post-treatment, but with greater pitch fluctuation. They spoke more tensely but more quietly, albeit with reduced loudness variability. Thus, only voice quality changed in the expected direction.
The study's hypothesis was therefore only partially supported, with consistent effects observed mainly for voice quality rather than across all predicted prosodic features.
According to the authors, this suggests that the 20-minute relaxation period is too short to induce consistently detectable or strong changes in prosody. In contrast, the enforced period of quiet idleness in the control group could potentially have induced stress or mental lethargy, which was absent from the other two groups. Similar changes have been suggested in previous research. This might account for the constricted or tenser voice in the control group.
Notably, the authors present these as exploratory interpretations, rather than objectively measured phenomena.
Study limitations
The study is limited by its small sample size, the lack of experimental corroboration of the suggested mechanisms, and the use of a relatively liberal statistical threshold that reflects its exploratory nature.
Speech prosody shows promise but remains exploratory biomarker
The results of this study show that speech prosody may serve as a sensitive yet exploratory biomarker to indicate the type of treatment effect produced and to monitor treatment. However, the findings also suggest that short interventions may not yield sufficiently strong or reliable prosodic changes for clear stress detection based solely on speech.
Further studies should use larger samples, a longitudinal design, and specifically assess gender-related effects, possibly comparing with other interventions. This could help develop rapid stress-relief interventions.
Download your PDF copy by clicking here.
Stress hormones disrupt the internal GPS system of the brain