Acoustic Correlates of Speech when Under Stress: Research, Methods and Future Directions

Abstract

The aim of this paper is to summarise the acoustical correlates of stress in speech by reviewing relevant empirical studies. Research into speech under stress faces several challenges including the difficulty in defining the concept of stress, limitations in collecting stressed speech experimentally and the problem of quantifying the type and level of stress induced, resulting in very limited comparability between various studies. In summarizing the previous work and evidence, these conceptual and methodological differences are addressed.
 Changes in respiration and muscle tension have repeatedly been evidenced to be physiological correlates of stress. Based on this, it may be expected that acoustic parameters such as fundamental frequency, intensity and speaking/articulation rate are affected by stress. In general, empirical findings support these predictions but inter-speaker variability must not be overlooked. It is argued that there is a need to move beyond these three parameters, taking into account for example, vowel formants and aspects of voice quality. It is also suggested that future research should aim to control the data gathering process in ways that are appropriate to allow for comparability of results between studies.

Individual differences in cortisol stress response predict increases in voice pitch during exam stress

“…Indeed, an increase in voice pitch is the longest and most commonly reported finding in previous studies examining speech under stress (reviewed in Giddens et al, 2013;Kirchhübel et al, 2011). However, many studies have failed to replicate this finding (e.g., Dietrich and Abbott, 2012;Hecker et al, 1968;Johannes et al, 2000;Streeter et al, 1983;Tolkmitt and Scherer, 1986;Van Lierde et al, 2009).…”

“…Studies examining the acoustic correlates of stress have produced widely mixed results (reviewed in Giddens et al, 2013;Kirchhübel et al, 2011). In large part these inconsistencies stem from differences in how stress is defined (Murray et al, 1996), difficulties related to inducing and measuring genuine stress responses in the laboratory (Berger et al, 1987;Kudielka and Kirschbaum, 2005;Ruiz et al, 1996;Scherer, 1979;Tolkmitt and Scherer, 1986), and substantial variation in the magnitude of the stress response across individuals (Berger et al, 1987;Johannes et al, 2000;Kudielka et al, 2007;Streeter et al, 1983).…”

“…Exam stress, the most significant source of stress experienced by students, peaking just prior to an academic examination (Robotham, 2008), offers an ecologically valid context in which to examine the effects of psychological stress on the voice. Psychological stress increases the activity of the hypothalamic adrenal (HPA) axis and circulating stress hormone levels, particularly cortisol (Dickerson and Kemeny, 2004;Kudielka and Kirschbaum, 2005), and is known to have multiple and apparent effects on the body including increased breathing rate, muscle tension, and changes in salivation rate, that may in turn affect vocal production (Kirchhübel et al, 2011;Scherer, 1981Scherer, , 1979.…”

“…Others report an increase in minimum voice pitch or a decrease in its standard deviation (F0 sd), with no systematic change in mean pitch (Park et al, 2011;Tolkmitt and Scherer, 1986). Previous studies also consistently indicate that voice pitch increases more in natural than artificially induced stress scenarios (Kirchhübel et al, 2011;Ruiz et al, 1996). It is possible that there are no reliable acoustic markers of psychological stress (Streeter et al, 1983).…”

“…It is possible that there are no reliable acoustic markers of psychological stress (Streeter et al, 1983). Alternatively, as has been suggested by many researchers, these inconsistencies in previous work may be tied to variation in the severity of the stress response, wherein vocal changes may be observed only in particular contexts (e.g., natural stress scenarios) or among individuals surpassing a critical threshold of stress (Beatty and Behnke, 1991;Johannes et al, 2000;Kirchhübel et al, 2011;Scherer, 1979). Despite this parsimonious possibility, only a handful of studies examining speech under stress have controlled for individual differences in stress levels (Hecker et al, 1968;Sigmund, 2006) or in participants' propensity toward anxiety (Beatty and Behnke, 1991;Tolkmitt and Scherer, 1986).…”