Cognitive stress, martial arts and human speech
Malcolm Evefors
COGNITIVE STRESS, MARTIAL ARTS AND HUMAN SPEECH Malcolm Evefors
Abstract
Results from multiple studies indicate fundamental frequency (F0) as a parameter of human speech that tends to increase with stress. This study compares voice recordings of experienced and novice martial artists, as practicing martial arts (MA) is speculated to improve stress management. The sample consisted of 10 participants, Qive experienced and Qive novices. Recordings were made in a normal state and a state of cognitive stress, induced by mental arithmetics and an auditive stressor.
Verbal stress response proved present for all participants. The experienced group showed lesser tendency to increase spread of frequency in the face of stress as compared to the novice group. The difference was conQirmed signiQicant by ANOVA. The results are similar to earlier Qindings on the effects of stress management training on verbal stress response. Further research is necessary to conQirm these results, considering homogeneity and sample size.
Sammanfattning
Resultat från Qlertalet studier indikerar att fundamental frekvens (F0) är en parameter i mänskligt tal som tenderar öka med stress. Denna studie jämför röstinspelningar av erfarna kampsportsutövare och noviser, eftersom kampsport har spekulerats förbättra stresshantering.
Urvalet bestod av 10 deltagare, fem erfarna och 5 noviser. Inspelningarna utfördes i ett normalt tillstånd samt i ett tillstånd av kognitiv stress, inducerat av mental aritmetik och en auditiv stressor.
Alla deltagare uppvisade verbal stressrespons. Den erfarna gruppen visade en mindre tendens att öka spridning av frekvens i en stressig situation, jämfört med noviserna. ANOVA bekräftade att skillnaden var signiQikant. Resultaten liknar de från tidigare forskning på effekten av stresshanterings-‐träning på verbal stressrespons. Ytterligare forskning är nödvändig för att bekräfta dessa resultat, med fokus på homogenitet och urvalsstorlek.
Stress is a constant research topic in many scientiQic Qields today. In the Qield of cognitive science, stress is frequently researched because of its numerous effects on cognition. Driven by neurological systems and hormone responses (Tsigos and Chrousos, 2002), stress can enhance as well as impair cognitive functions. Among other functions, stress has proven to affect attention, decision making and memory (Mendl, 1999). Over time stress might even come to change the structure of the brain in the hippocampal region (McEwen and Sapolsky, 1995). While the effects of stress on neurological and cognitive functions are important to understand, it’s also an important direction of research to study methods of managing stress.
In the area of linguistics, researchers have been studying stress by analyzing
speech samples from stressful situations. It is not hard to see how greater
understanding about how stress speciQically affects speech production could have
multiple applications: Perhaps an automated system being able to accurately estimate the gravity of a emergency call, or diagnostics able to determine exactly the stress coping capabilities of an individual based simply on conversation.
Some studies have been dedicated to examining what parameters of the speech signal are especially prone to be affected by stress. Results from multiple studies indicate that fundamental frequency (F0) is one such parameter. Streeter, Macdonald, Apple, Krauss and Galotti (1983) examined tape recordings of a conversation between a system operator at a major electricity company and his superior. The conversation started at the time of some complications, which an hour later would cause the great New York blackout of 1977. Analysis saw a signiQicant increase in F0 for the superior with increased stress, while the system operator’s levels decreased.
Fundamental frequency (F0) is the base frequency of a periodic waveform, which in speech is signiQied by the vibrations caused by the vocal folds. The idea of F0 as a vocal indicator of stress is further strengthened by the results from Streeter et al.
(1983), where listeners were able to identify stress levels from the recordings with a positive relation to the average pitch. That the system operator didn’t show an increase in F0 has been argued to be because the operator was not as stressed as his superior, or to implicate differences between individuals in how parameters are affected.
Studying speakers in life-‐threatening scenarios, Gramatica, Ruiz and Legros (1992) examined speech from a cockpit voice recorder (CVR). The CVR was that of a plane that came down with serious malfunction and ended in a fatal crash. Analysis saw a signiQicant increase in mean fundamental frequency (F0mean) as the situation grew more stressful for the pilots. The study also examined recordings of a participant in a simulated Qlight experience, where an signiQicant increase in F0mean was indeed observed, but not to the same extent as in the real recording.
A study by Absil et al. (1995) also examined speech from a natural situation (CVR sample) as well as a recording from a laboratory setting (Stroop test). As in Gramatica et al. (1992), signiQicant increases in F0mean were observed for both situations; the natural situation seeing a greater increase than the laboratory one.
Looking more directly into the laboratory setting, a study by Hecker, Stevens, von
Bismarck and Williams (1968) looked to induce stress by making participants
solve math problems under time pressure. Comparison of voice recordings from
before and after being subjected to the stress condition showed increase in F0mean for some of the participants while some showed a decrease.
Tolkmitt and Scherer (1968) subjected participants to what was classed as emotional and cognitive stress. Participants were classed a high anxiety, low anxiety or anxiety deniers. Results indicated no effect on mean F0 for any group, but an increased F0 Qloor was observed for participants classed as high anxiety or anxiety deniers.
In summary, research generally points to increase in F0mean as a verbal stress response in humans, but with varying levels between individuals and largely dependent on the type and intensity of stress induced. On the matter of the physiological mechanisms behind the effect, this is discussed in a recent review:
Vocal Indices of Stress: A Review. On a general level, stress might affect breath rate and air Qlow as well as heart rate and blood pressure, which in turn affects different constructs, essentially the vocal folds. The hypothalamic-‐pituitary-‐adrenal axis (HPA) and the autonomic nervous system (ANS) are known to be responsible for these kind of stress responses. It is also speculated that the parasympathetic branch of the ANS triggers nerve responses in the laryngeal and facial areas, affecting muscle tension and in extension speech production (Giddens, Barron, Byrd-‐Craven, Clark and Winter, 2013).
Based on previous research, Jessen (1997) compared the verbal stress response of a group that had stress management training with a control group with no training.
The two groups were subjected to two stress conditions, one focused on cognitive stress and the other physical stress. Both groups showed signs of increased F0mean values and standard deviation (F0sd) for both the conditions. However, for the cognitive stress condition, the stress management group showed signiQicantly less increase compared to controls. Jessen argues this to be indicative of better stress coping/management strategies for that group.
Jessen (1997) argues that the perceived threat level to the person is relevant to whether a subject will reach such levels of stress as to bring signiQicant F0 increase.
This appears an okay Qit considering prior research; high threat scenarios such as
aviation accidents (Gramatica et al., 1992, Absil et al., 1995) or the great New York
blackout of 1977 (Streeter et al., 1983) have seen generally high F0 increase, while
low threat scenarios such as laboratory experiment participation (Hecker et al.,
1968, Tolkmitt & Scherer, 1986) have yielded lower or insigniQicant results.
To impose the medium threat level to which the study results are attributed, a professional tie in was made. Both participant groups consisted of police ofQicers, with the stress management training as well as the stress condition engagements being police work related. During participation, police academy teachers and senior ofQicers were present to make the participants feel a potential threat to their career or employment.
However, the study differs on another point from prior research: Whereas the main body of stress inducing tasks used in earlier research has been focused on “simply”
performing mental arithmetics following visual instructions, this study brought in an auditive aspect. The participants were, at the same time as they were performing a mathematical test, made to listen to a hostage recording and try to remember events. Not only does this mean a more stressful situation for the participants because of the dual task assignment; it’s a fair assumption to make that the audio, being from a hostage situation, makes for a disturbing factor. Note that this is not an argument against the weight that perceived threat level might have in creating verbal stress response. However, this study considers the possibility that the addition of dual task assignments as well as an auditory stressor might on their own be able to induce cognitive stress potent enough to force signiQicant F0 increase.
Inspired by the results observed by Jessen (1997), the purpose of this study is to examine if similar effects can be found for another type of population than the policemen of that study. For an accessible population with some relation to stress management, this study will be looking at practitioners of martial arts (MA). While more research still needs to be done in the MA Qield, some indications have come up, suggesting that the awareness and mindfulness approach to training might improve stress management. In the review Martial arts: Mindful exercise to combat stress, the conclusion is expressed: “The practice of MA appears to provide several beneQits for the management of chronic stress”. (Naves-‐Bittencourt et al., 2015, p.
46) While this study does not focus on chronic stress per se, we assume to some extent a transfer effect between management of chronic stress and more short term cognitive stress.
In addition to analyzing F0mean and F0sd, range of fundamental frequency
(F0range) will also be included in this study. This is based on the aforementioned
results obtained by Tolkmitt and Scherer (1986), observing increase in F0 Qloor for
anxious participants as compared to the not anxious participants. Tolkmitt and
Scherer argue that the increased Qloor values might be because of increase in
muscle tension. This works in line with the speculations made by Giddens et al.
(2013) on the physiological mechanisms behind the verbal stress response.
Despite this, the F0 Qloor variable has been left unexamined after Tolkmitt and Scherer (1986). We argue that if F0 Qloor is increased, this would make for a smaller F0range. It is possible that by earlier focus on F0mean and F0sd, this effect of stress on the range of the frequencies has gone unnoticed.
To control for eventual differences in receptibility to the stressors, this study also includes a post-‐participation survey to be able to control for any relationship between the participants perceived stress level and observed verbal stress response.
In sum, this study aims to answer the question: Does stress management training affect the verbal stress response to induced cognitive stress? Based on the earlier research, the hypothesis is that stress management training leads to a lesser verbal stress response. On the level of each of the measured F0 variables, the hypothesis is expressed as:
1. Stress-‐induced difference in F0mean will be lower in the presence of stress management training.
2. Stress-‐induced difference in F0sd will be lower in the presence of stress management training.
3. Stress-‐induced difference in F0range will be lower in the presence of stress management training.
Method Participants
A convenience sample of ten voluntary participants from a martial arts organization was selected. To ensure homogeneity in regard to physical and mental training regime, all participants were practitioners of the same MA. The original plan was to, also considering homogeneity of results, limit the study to participants of one gender. However, working with a convenience sample from a limited population, this was not possible.
Based on how many years the participants had been practicing, they were categorized as belonging to one of two groups: the novice group or the experienced group. For the novice group, criteria for inclusion was having practiced for a maximum of 2 years. For the experienced group, criteria were to have practiced for at least 5 years.
The novice group consisted of 5 participants (3 female, 2 male). The average age of
the participants in the group was 34.4 years (S.D 16.3) and they had on average
been practicing the MA for 1.5 years (S.D .35). The experienced group also consisted of 5 participants (5 male). The average age of the group was 21.2 (S.D . 83) and they had been practicing for 12.6 years on an average (S.D 3.9).
Ethics
The study as a whole and in particular everything regarding the stress condition follows the rules and guidelines of the CODEX as stated by Vetenskapsrådet. Data collection and storage follows the rules as stated in Personuppgiftslagen (PUL 1998:204). The wellbeing of the participants was considered carefully when constructing the stressors of the tasks. Considering the short time exposed to the stressors as well as the highly cognitive nature of the stress, stressors were deemed ethical and approved. While introduction and instructions were crafted to enforce performance pressure, the participants were relieved of this during the debrieQing, as to not leave any participants with any long time anxiety. To maximize degree of anonymity, the particular martial art practiced by the participants as well as the martial arts organization they belong to, has been left out.
Apparatus and materials
The internal microphone of a laptop computer (Model: MacBook 7.1) was used for recording. Recordings were made in 4 rooms because of accessibility issues. To control for eventual static or noise, test recordings were made in the rooms prior to the actual recording sessions. Software used for recording as well as analyzing results was Praat version 6.0.16. Auditory stimulus was presented via on-‐ear headphones (Model: Creative HS 450).
In the study by Jessen (1997), cognitive stress was induced by combining a task with mathematical focus with an audio based task focused more on remembering.
For the purpose of this study, it was decided to not utilize the same kind of audio based task. Using an actual recording could prove too much for the participants to handle emotionally and a was deemed unethical. Using a staged recording (i.e from a movie) was considered, but was also dismissed: it could also end up being too much to handle, or on the opposite end recognized for being staged. To be able to keep the dual task aspect from the earlier study, two arithmetic tasks were chosen to be performed alternately. To still be able to include the aspect of the auditive stimulus, it was decided to use an auditive stressor.
A pre-‐study was carried out to determine the auditive stressor. In a study from
2007, Cassidy and MacDonald examined background noise, music and their effects
on performance in a variety of cognitive tasks. Compared to silence, background
noise and music with high or low arousal were all detrimental to performance.
High arousal (high information load, negative affect) music proved to be the most detrimental and thus a prime stimulus choice for this study. The primary purpose of the stimulus was to serve as a distractor, detrimental to performance, to place greater pressure and cognitive load on participants. Four participants unrelated to the the main study were subjected to three music items. The three items up for selection were Requiem by György Ligeti, Atmospheres by György Ligeti, and The Beast by Jóhann Jóhannsson. György Ligeti (1923-‐2006) was an avant-‐garde composer, known for composing works in an atonal fashion that he himself named as micropolyphony. He is probably best known from excerpts of his music being used in the soundtrack to the movie 2001: A Space Odyssey. Jóhann Jóhannsson is a contemporary composer, connected to electronic and drone music. All items were classiQied as being of high arousal type. Participation included passive listening as well as listening while performing mental arithmetics. Participants were asked to report which item was the most disturbing. Atmospheres was nominated unanimously and chosen as stimulus.
The text to read for the recording was the Swedish version of the text The North Wind and the Sun. Using a predetermined text was chosen to, on a general level, make sure the speech samples were as similar as possible. Reading a predetermined text controls for things such as accentuation and tonal variance between questions and statements. The text speciQically was chosen because of the format allowing for overview and easy analysis, as well as it being established as a text for voice analysis, both internationally as well as nationally. (Handbook of the the International Phonetic Association, 1999)
The Qirst arithmetic task was a matter of multiple digit additions. Given a three digit number, the task is at Qirst to sum together the value of each digit. After this, the sum is added to the original number, with the result being the answer. Example:
‘123’ is the given number. Each individual number is added together (1+2+3=6), yielding a sum which is meant to be added to the original number (123+6=129). In this case 129 is the correct answer. There was a total of 60 tasks available to the participants. The numbers were selected by randomly generating a series within the interval 100-‐999. Numbers that ended up even hundreds (100,200,300…) were replaced to make sure there would be no digit additions with a single digit and two zeroes. The list of tasks can be found in Appendix A.
The second arithmetic task was the classic task of the serial sevens. While the the
efQiciency of the serial sevens test in assessing mental functions are being
discussed, it's still considered cognitively demanding, relying on attention and
concentration (Ganguli et al., 1990). The task is to count downwards in steps of
seven from a given number, typically 100. In the case of this study, the Qirst number was decided to be higher. This was because of the amount of addition tasks available for the participants to attempt, making sure that reaching 0 in the serial task was made impossible. Arbitrarily, 732 was chosen as the starting number. In the case of a participant losing track of the serial counting, after a 10 second period the researcher would let the participant know what their most recent guess had been, switched to their second most recent guess if track was lost again. This was to ensure continued exposure to both kinds of tasks.
For the post-‐participation survey, there were only two items. One relating to perceived mental strain regarding the mathematical aspect and one relating to perceived mental strain regarding the auditive aspect. The items were set out to state that the aspect had been straining, with options allowing the participants to select to which extent they agreed with the statement. The options allowed for not agreeing at all, agreeing to a small extent, agreeing to a large extent or agreeing fully. The survey itself can be found in Appendix B.
Procedure
Before participation, informed oral consent was obtained. The participants were informed about how information about them and their participation were to be handled anonymously. They were informed that participation would include sound recordings and that they were to perform mental tasks while subjected to background noise through headphones. Participants were not informed about the actual purpose of the study, which instead was presented as being “investigating performance on mental tasks while subject to background noise”. This was to avoid the participants consciously altering their voice from being aware it might affect their results.
After the brieQing, the Qirst part of the session was calibrating the sound level of the auditory stimulus. The participants were allowed to listen to a few seconds of the stimulus on loop (5 seconds x 3 repetitions) while they were instructed to adjust the volume to as high a level as possible without reaching any levels of pain or discomfort. Participants were told that in case any sounds became painfully loud in any later stages, they would be allowed to lower the volume to a manageable level.
This was an attempt to lower the probability of any participants quitting before completing the assignment, in case the stimulus would prove too loud over the course of time. It was also reasoned to possibly make the participants go into the assignment with a higher initial volume, knowing that they could adjust the volume if absolutely necessary.
The calibration was followed by the Qirst recording. Participants were presented with the text to read and instructed to, after internally counting to three, read it out loud as naturally as possibly. The internal counting was a measure taken to reduce noise of paper rustling during the Qirst utterances.
After reading the text out loud, participants were given instructions about the math tasks. The participants were instructed to solve the Qirst addition before verbally announcing the Qirst step in the serial countdown, after which to continue performing the tasks in that order. The addition tasks were printed on paper and the answers supposed to be written by pen. The answers to the serial task were to be given verbally and written down by the researcher. They were instructed that their performance would be graded depending on the amount of addition tasks they attempted, the amount of correct addition solutions and the amount of correct serial countdowns, so that they could not neglect one of the tasks for the other one.
They were also made aware that the they would only have a certain amount of time to complete the tasks. No information was given as to what would happen if they lost track during the serial counting task. After receiving the full instructions, the participants were equipped with the calibrated headphones, they were told to start and the music would start playing. The participants were given 10 minutes to complete as many tasks as possible. An abundance of addition tasks had been prepared to make sure there was no risk of any participant Qinishing before 10 minutes.
After ten minutes, the auditive stressor was turned off, the participants would be told to put down their pencil and remove the headphones. Instructions were then given for the second recording, mirroring the instructions from the Qirst. There were no direct indications that the text would be the same as for the Qirst recording, to try and ensure that the reading would be as natural as the Qirst time.
Directly after the second recording the participants were given the post-‐
participation survey. Upon completing the survey the participants were debriefed and informed about the true purpose of the study. From introduction to debrieQing, each session had an average length of 25 minutes.
Each participant’s post-‐participation survey responses were transformed to variables. The option not agreeing at all was coded as 1, agreeing to a small extent as 2, agreeing to a large extent as 3 and agreeing fully as 4.
Using the Praat software, visual checks for abnormal values in the recordings were
carried out. Abnormal values were found in three instances and subsequently
removed. F0mean, F0sd as well as F0range values were then extracted for each recording for each participant. The differences for the three variables between recordings were calculated; yielding a F0mean difference, F0sd difference and a F0range difference for each participant. To compare results on the group level, three separate one-‐way ANOVAs were then carried out. Dependent variables were difference in F0mean, difference in F0sd and difference in F0range, with group, reported math-‐related stress and reported sound-‐related stress as independent variables respectively.
Results
In this section, in separate tables, we present the data extracted from the sound Qiles regarding variables F0mean, F0sd and F0range. For each variable, a table presents the mean difference from the relaxed to the stressed state for both groups. A table of the reported math-‐ and sound-‐related stress ratings is also presented. ANOVA results for each dependent and each independent variable are then presented.
Table 1. Mean (and standard deviation of) F0mean difference from relaxed to stressed state, for the novice group and the experienced group.
Group Mean F0mean difference (In Hz) S.D (In Hz)
Novice -‐5.24 10.53
Experienced -‐0.66 5.16
The F0mean results on (Table 1) show that both groups tend to decrease in F0 from the relaxed to the stressed state. Compared to the novice group, the experienced group displays a lesser mean difference as well as a lower standard deviation. The differences, however, are limited in being just below 5 Hz (Mean), and just over 5 Hz (S.D).
Table 2. Mean (and standard deviation of) F0sd difference from relaxed to stressed state, for the novice group and the experienced group.
Group Mean F0sd difference (In Hz) S.D (In Hz)
Novice 10.09 6.47
Experienced .56 3.18
For F0sd difference (Table 2), there are larger differences. With a more substantial increase just over 10 Hz, the novice group shows a tendency to increase standard deviation from the relaxed to the stressed situation. The experienced group shows close to no difference, with a value just over .5 Hz.
Table 3. Mean (and standard deviation of) F0range difference from relaxed to stressed state, for the novice group and the experienced group.
Group Mean F0range difference (In Hz) S.D (In Hz)
Novice -‐4.97 27.16
Experienced 20.48 77.8
The results regarding F0range difference (Table 3), indicate largely different responses from the two groups. In the stress condition, the novice group shows tendency to decrease slightly in F0range, while the experienced group displays a large increase. However, large standard deviations for both the groups are indicative of a greater spread of individual values within the groups.
Table 4. Number of math- and sound-related stress reports, for each group. (1= not agreeing at all, 2= agreeing to a small extent, 3= agreeing to a large extent, 4=
agreeing fully, where that which is being agreed to is feeling stressed)
Math-‐related stress Sound-‐related stress Group 1 2 3 4 1 2 3 4
Novice 0 2 2 1 0 2 0 3 Experienced 0 2 2 1
0 3 0 2
All participants agreed that they had been stressed to some extent by participating
(Table 4). Comparing the groups, reports are quite similar; The only differences can
be found in sound-‐related stress, where the novice group has one more full
agreement as compared with the extra agreeing to small extent for the experienced group. Note that the lack of reported 1’s limits the ANOVA with math-‐related stress as independent variable to three groups. The lack of 1’s and 3’s limits the ANOVA with sound-‐related stress as independent variable to two groups.
For all the ANOVAs carried out, F0mean difference, F0sd difference and F0range difference were set as dependent variables.
One-‐way ANOVA, group as independent variable:
With group as independent variable, analysis showed insigniQicant results for both F0mean and F0range differences [F(1,8) = .762, p = .408] [F(1,8) = .477, p = .509].
However, for F0sd difference, results indicate signiQicant difference [F(1,8) = 8.746, p = .018]. This signiQies the experienced group showing signiQicantly less increase in standard deviation than the novice group.
One-‐way ANOVA, math-‐related stress as independent variable:
When running the ANOVA using reported math-‐related stress as independent variable, results were insigniQicant for all dependent variables; F0mean difference [F(2,7) = .626, p = .562], F0sd difference [F(2,7) = .513, p = .619] and F0range difference ([F(2,7) = .185, p = .835]).
One-‐way ANOVA, sound-‐related stress as independent variable:
Using the third independent variable, reported sound-‐related stress, results were once again insigniQicant for all three variables; F0mean difference [F(1,8) = .005, p
= .943], F0sd difference [F(1,8) = .065, p = .805] and F0range difference ([F(1,8) = . 017, p = .898]).
Discussion
It was assumed that the use of dual tasks as well as the auditive stimuli as stressors would be sufQicient in inducing verbal stress response, without the pressure of stress related to profession or career. This seems accurate as the results indicate that the induced cognitive stress affected all measured variables to some extent.
This was true for both groups.
The groups did not differ signiQicantly in responses when looking at F0mean and
F0range: This is not in line with the hypothesis of the study. However, for the F0sd
parameter, there was a statistically signiQicant difference between groups. Indeed,
the increase in standard deviation for the novice group signiQies that as stress
increases, the frequencies in a sample are generally distributed further from the
mean. The signiQicantly smaller effect for the experienced group then indicates speech signals with a general distribution closer to the mean. This means that when faced with cognitive stress, the novice group show a tendency to increase spread of frequency, while the experienced group is able to better maintain the spread from a relaxed state. The F0sd results follow the hypothesis in that the experienced group would show less verbal stress response than the novice group.
The results regarding the F0range do not support the hypothesis, with the novice group showing less of a difference between conditions than the experienced group.
It is a possibility that the effect of increased F0 Qloor (Tolkmitt & Scherer, 1986) that the study of this variable was motivated by, only takes place in subjects with anxiety or stress coping issues. With the wide spread in these results however it should still be left to future research to study this further.
As difference in F0mean has been the main indicator of verbal stress response in prior research, it is peculiar that between-‐groups comparison shows no signiQicant difference for this variable. It's important to remember that while the universal trend of increases was not observed, the tendency for decreases found in this study is also classed as a sign of verbal stress response. A signiQier for insufQicient stress would rather have been an indifference between recordings.
While the hypothesized results were only seen for one of the three measured variables, we argue that the F0sd data should be seen as indicative that stress management training does reduce verbal stress response. The lack of difference between groups regarding F0mean is considered dependent on the unknown potency of effects of MA as stress management training, as well as the laboratory nature of the setting. More in-‐depth speculations on these effects of MA as stress management training is left to further research that may target that subject speciQically.
While we do argue for the indications seen in this study, they need to be conQirmed
through additional research. For this, there are a number of considerations to keep
in mind. In general, to be able to conQirm the results of this study, it is necessary for
future research to include a larger and more homogeneous sample. Controling for
differences in age and gender, which might have had unexamined effects on the
results of this study, is recommended. This study is also subject to the multiple
testing problem, where multiple analyses are being run on the same participants. A
signiQicance level of p<.05 was used throughout, where it would have been
appropriate to utilize Bonferroni correction to adjust for the multiple testing. We
advice to keep these kinds of adjustments in mind or limit studies to more singular
analysis. Aside from these major points, it might be worth relying on standardized stressor material to be able to more reliably control for varying effects of stressors between participants.
In summary: From all that has been observed in this study, greater experience with
stress management training seems to correlate with a lesser verbal stress response
as compared to being novice. This is expressed in the form of showing less
tendency to increase F0sd with increased cognitive stress. These results are
tentative however and need to be conQirmed through additional studies. Research
delving further into this area should also include larger sample sizes as well as
taking homogeneity into account.
References
Absil, E., Gramatica, B., Harmegnies, B., Legros, C., Poch, D., & Ruiz, R. (1995, September). Time-‐
related variabilities in stressed speech under laboratory and real conditions. Paper presented at the ESCA-‐NATO Workshop on Speech under Stress, Lisbon, Portugal.
Cassidy, G., & MacDonald, R. A. (2007). The effect of background music and background noise on the task performance of introverts and extraverts. Psychology of Music, 35(3), 517–537.
Ganguli, M., Ratcliff, G., Huff, F. J., Belle, S., Kancel, M. J., Fischer, L. (Kuller, L. H.) (1990). Serial sevens versus world backwards: a comparison of the two measures of attention from the MMSE. Journal of
Geriatric Psychiatry and Neurology, 3(4), 203–207.Gonzalez, F. (2009, January 28). Atmospheres-‐Gyorgy Ligeti [Video Qile]. Retrieved from https://www.youtube.com/watch?v=aI0P1NnUFxc
Gramatica, B., Ruiz, R., & Legros, C. (1992). ModiQication de la fréquence fondamentale de la voix des pilotes: Incidents réels et simulés. Le Journal de Physique IV, 2(C1), 335–338.
Hecker, M. H., Stevens, K. N., von Bismarck, G., & Williams, C. E. (1968). Manifestations of Task-‐
Induced Stress in the Acoustic Speech Signal. The Journal of the Acoustical Society of America, 44(4), 993–1001.
International Phonetic Association. (1999). Handbook of the International Phonetic Association: A
guide to the use of the International Phonetic Alphabet. Cambridge University Press.Jessen, M. (1997). Phonetic manifestations of cognitive and physical stress in trained and untrained police ofQicers. International Journal of Speech Language and the Law, 4(1), 125–147.
McEwen, B. S., & Sapolsky, R. M. (1995). Stress and cognitive function. Current Opinion in
Neurobiology, 5(2), 205–216.Mendl, M. (1999). Performing under pressure: stress and cognitive function. Applied Animal
Behaviour Science, 65(3), 221–244.Mifalco (2012, April 23). Ligeti -‐ Requiem (1965) [Video Qile]. Retrieved from https://www.youtube.com/watch?v=wawSCvuGj4o
Naves-‐Bittencourt, W., Mendonça-‐de-‐Sousa, A., Stults-‐Kolehmainen, M., Fontes, E., Córdova, C., Demarzo, M., & Boullosa, D. (2015). Martial arts: mindful exercise to combat stress. European
Journal of Human Movement, 34, 34–51.Soundtracked Universe (2015, October 3). Sicario Soundtrack -‐ The Beast [Video Qile]. Retrieved
from https://www.youtube.com/watch?v=sN9gcEZMZO8
Streeter, L. A., Macdonald, N. H., Apple, W., Krauss, R. M., & Galotti, K. M. (1983). Acoustic and perceptual indicators of emotional stress. Journal of the Acoustical Society of America, 73(4), 1354–
1360.
Tolkmitt, F. J., & Scherer, K. R. (1986). Effect of experimentally induced stress on vocal parameters.
Journal of Experimental Psychology: Human Perception and Performance, 12(3), 302–313.