Researchin Nursing & Health, 1992, 15, 379-309
Validity and Reliability of Nonverbal Voice Measures as Indicators of Stressor-Provoked Anxiety Barbara F. Fuller, Yoshiyuki Horii, and Douglas A. Conner
To determine the validity and reliability of vocal jitter, tenseness, and pitch as measures of stressor-provoked anxiety, 88 women representing three coping styles (Highly Anxious, Truly Low Anxious, and Repressor) phonated vowels, reported anxiety, and provided measures of muscle tension, heart rate, and sweating 2 weeks before, the day before, and l week after stressful exams. Greater self-reported anxiety, sweating, and heart rate on the day before versus the other occasions confirmed the day before was the most stressful occasion. Results suggest excellent validity and reliability for Jitter as an indicator of stressor-provoked anxiety. @ 1992 John Wiley & Sons, Inc.
Lowery (1987) presents a useful model for nursing research concerning stress. This model encompasses the stressor (life events provoking stress-related responses), mediators (social support, coping behaviors, coping styles), responses (psychological and biological reactions), and outcomes (health, illness), together with their reciprocal influences. A commonly measured psychological reaction is anxiety (Robinson, 1990; Scott, Oberst, & Dropkin, 1980). Biological reactions, termed physiological arousal, include increased blood pressure, heart rate, respirations, sweating, and skeletal muscle tension (Cannon, 1929; Guyton, 1986; Robinson, 1990). The use of these biological reactions to indicate anxiety in nursing assessments is hindered by several limitations. First, measures of physiological arousal vary greatly among individuals (Hugdahl, 1981; Lacey & Lacey, 1958). Second, only weak relationships exist among physiological measures or between groups of physiological measures and groups of psychological measures of anxiety (Morrow & Labrum, 1978; Nesse et al., 1985; Tyrer & Lader, 1976). Third, pathology (Guyton,
1986; McCance & Heuther, 1990) and medications (Swonger & Matejski, 1991), alter physiological measures. Last, methods for obtaining physiological measures can be intrusive. Psychological measures of anxiety also have limitations. The clinical usefulness of self-reported anxiety is hampered by its amenability to intentional or unintentional misrepresentation (Fagin, 1987). Unintentional misrepresentation of self-reported anxiety can stem from certain coping styles (i.e., styles of perceiving, evaluating, and expressing stressor-related anxiety). One coping style, termed Highly Anxious, is a hypervigilance concerning anxiety-related thoughts (Lazarus & Folkman, 1984) that results in high levels of both self-reported anxiety and physiological arousal (Weinberger, 1990). Another coping style is Repression. Repressors exhibit elevated thresholds for recognizing disturbing feelings of anxiety or stress (Krohne & Rogner, 1982) and prefer socially desirable attributes (Bemhardson, 1%7; Byme, 1964). They consistently address anxiety by minimizing cognitions about it and expressions of it (Lazarus & Folkman, 1984). That their level of physiological
Barbara F. Fuller, PhD, RN, is a professor, School of Nursing; Yoshiyuki Horii, PhD, is a professor, Department of Speech Science and Communication Disorders, and Douglas A. Conner, PhD, is a research fellow, School of Nursing; all are at the University of Colorado. This study was supported by Grant NR01468 from the National Center for Nursing Research, National Institutes of Health, to the first two authors. The authors are grateful to Dr. Carol Vojir for statistical advice. This article was received on September 6, 1991, was revised, and accepted for publication April 27, 1992. Requests for reprints can be addressed to Dr. Barbara F. Fuller, Campus Box C-288, University of Colorado Health Sciences Center, 4200 East Ninth Ave., Denver, CO 80262.
8 1992 John Wiley & Sons, Inc. CCC 0160-6891/92/050379-11
380
RESEARCH IN NURSING 8. HEALTH
arousal is similar to, or greater than, levels of Highly Anxious subjects repudiates their social expressions of low anxiety (Asendorpf & Scherer, 1983; Fuller, 1992; Jamner & Schwartz, 1986; Weinberger, 1990). Truly Low Anxious individuals report low levels of anxiety and their physiological arousal is lower than that of Highly Anxious and Repressor subjects (Asendorpf & Scherer, 1983; Weinberger, 1990). The lack of strong correlations between selfreported anxiety and either physiological arousal measures (Nesse et al., 1985) or nursing assessments (Kruger, 1983) may have resulted from combining findings from subjects with different coping styles. A noninvasive, interval-level measure of anxiety would facilitate clinical nursing research with patients who are unable to provide reliable self-reports (e.g., confused elderly). Ideally, such a measure should not differ across coping styles nor be open to intentional misrepresentation. Potential stressor-induced alterations in acoustic voice properties seemed promising as such a measure. Combinations of acoustic properties can differentiate among infants’ “hungry,” “fussy,” and “pain” cries (Fuller, 1991). Vocalizations are composed of combinations of simple sound waves. Each wave has a characteristic frequency and amplitude. Frequency is reported in cycles per second (Hz). Amplitude (the energy contained in the waveform) is measured in decibels (dB). A voice spectrum is a plot of the frequency and amplitude of all the simple waves in a vocalization. The lowest frequency of simple sound waves, termed the fundamental fiequency, is pitch. Variability in the durations of individual sound waves of pitch is termed jitter and is perceived as roughness. Tenseness indicates the relative amount of sound energy in the higher versus lower frequencies in the spectrum (Laver, 1980). Tense voices sound strident or metallic. Sensitive listeners can perceive larger variations in these acoustic properties. Laboratory analysis is required to detect smaller variations. Scherer (1986) theorized that stressor-provoked physiological arousal affects the voice in two ways. First, the arousal increases the tautness of laryngeal and vocal fold muscles which increases pitch and may also increase jitter. Second, the arousal constricts the superior larynx and pharynx and tenses the remaining supraglottal airway muscles which, collectively, elevate the ratio of energy in the higher to lower voice frequencies, thereby increasing tenseness. Accordingly, the major aim of this study was to ascertain the validity of one or more nonverbal voice characteristics (pitch, jitter, and/or tenseness)
as a noninvasive measure of stressor-provoked anxiety. This aim was implemented by (a) determining whether the values for each voice measure differed between an occasion of greater stressorprovoked anxiety (the day before the subject’s graduate comprehensive examination) and two occasions of lesser stressor-provoked anxiety (2 weeks before and 1 week after the exam), and (b) determining if any difference in voice measure values occurred for subjects as a function of coping style: Truly Low Anxious, Highly Anxious, and Repressor. Because the system for classifying coping styles (Weinberger, S c h w a , & Davidson, 1979) and the studies that validated this system used only laboratory stressors, an ancillary study aim was to determine the psychological and biological responses of subjects to a naturally occumng stressor by coping style. The research hypotheses were: 1. Self-reported state anxiety for all subjects will be greater on the day before than on either of the other two less anxiety provoking occasions. 2. Heart rate, palmar sweating, and forehead muscle tension for all subjects will be greater on the day before than on either of the other two occasions. 3. The self-reported anxiety of Repressors will: (a) not differ from, or be lower than, that of Truly Low Anxious subjects; and (b) be lower than that of Highly Anxious subjects. 4. Highly Anxious and Repressor subjects will demonstrate greater physiological arousal than Truly Low Anxious subjects. 5 . The vocal measures of stressor-inducedanxiety will: (a) be greater on the day before than on either of the other two occasions, (b) show variation by stressor condition regardless of coping style, and (c) not differ across coping styles. 6. The short-term (2 min) stability of pitch, jitter, and tenseness measurements will be excellent (r 3 .85).
METHOD
Subjects The convenience sample consisted of 88 healthy female postbaccalaureate health care students from a university in the Westem U.S. who signed an informed consent, had no known pathology, and reported taking no medications except occasional aspirin for headache. Women were used because
NONVERBAL VOICE MEASURES / FULLER ET AL.
(a) the potential subject pool contained few men, and (b) voice data for men and women cannot be combined due to gender differences in pitch. Ages ranged from 22 to 58 years, with a mean of 3 1.9 & 7.4 (SD)years. Thirty-one women were classified as Truly Low Anxious, 31 as Highly Anxious, and 26 as Repressors according to the method described below.
Measures Combinations of the normative median score (13) on the Taylor Manifest Anxiety Scale (MAS) (Taylor, 1953) and the normative upper quartile scores (18 and above) on the Marlowe-Crowne Social Desirability Scale (SDS) (Crowne & Marlowe, 1964) were used, per the method delineated by Weinberger and coworkers (1979), to classify persons as: (a) Truly Low Anxious (MAS score < 14 and SDS score < 18), (b) Highly Anxious (MAS score > 13 and SDS score < 18) or (c) Repressor (MAS score of < 14 and SDS score > 17). Studies concerning construct validity of the Weinberger method in men (Weinberger et al, 1979; Asendorpf & Scherer, 1983) showed that Repressors responded to stressful laboratory tasks with greater palmar sweating and forehead muscle tension, but lower self-reported anxiety, than did Truly Low Anxious subjects. Highly Anxious subjects reported greater state anxiety than did Truly Low Anxious subjects or Repressors. Although one stressful laboratory task did not affect heart rates differently across coping styles in one study (Weinberger et al., 1979), Highly Anxious and Repressor subjects of both sexes responded to another stressful laboratory task with faster heart rates than did Truly Low Anxious subjects (Asendorpf & Scherer, 1983). A similar stressful laboratory task also prompted a greater elevation in the diastolic blood pressure of Repressors than of Truly Low Anxious or Highly Anxious subjects of both sexes (Kiecolt-Glaser& Greenberg, 1983). In a recent study (Fuller, 1992), female Repressors reported less state anxiety and evidenced lower vagal cardiac tone (as measured by respiratory sinus arrhythma and interbeat interval variability) in response to a naturally occumng stressor than either Truly Low Anxious or Highly Anxious females. Truly Low Anxious subjects reported less state anxiety and showed greater vagal cardiac tone than did Highly Anxious subjects. Content validity of the SDS as a measure of affect inhibition and repressive defensiveness (Crowne & Marlowe, 1960, 1964; Millham &
381
Jacobson, 1978) was established by a panel of 10 experts. Correlationsof the SDS with anxiety scores were between -.2 and -.4 (Millimet, 1970). No gender differences were evident in normative scores; stability (> 1 month) was 3 8 , as was internal consistency (Crowne & Marlowe, 1964). The internal consistency of the SDS in this study was 30. Anxiety was measured with the Taylor Manifest Anxiety Scale (Taylor, 1953) and Speilberger’s State Anxiety Inventory (Spielberger, Gorsuch, & Lushene, 1983). Extensive reliability and validity data have been reported for both (Buss, 1955; Hoyt & Magoon, 1954; Spielberger, 1975; Taylor, 1953). No gender difference emerged in normative scores for either tool. Alpha coefficients were 0.85 for the Manifest Anxiety Scale and 0.96 for the State Anxiety Inventory in this study. Stressor-induced physiological arousal was measured with heart rate, palmar sweating, and forehead muscle tension. The validity and reliability of these responses are widely accepted (Cannon, 1929, Guyton, 1986; Morrow & Labrum, 1978; Nesse et al., 1985; Robinson, 1990; Tyrer & Lader, 1976). Physiological measures were simultaneously recorded on moving chart paper using a Grass Model 79D polygraph. For heart rate, Q-waves were obtained from plate electrodes attached on the right leg and forearm. These waves were converted by the tachograph channel into beatshin. Palmar sweating was recorded inversely as Kohms of skin resistance. Resistance was measured using two Ag/AgCI electrodes, filled with isotonic paste, attached to opposite dorsal and ventral surfaces of the middle finger of the nondominant hand. The greater the palmar sweating, the lower the resistance. Forehead skeletal muscle tension, expressed in microvolts, was measured using two stainless steel 9-mm cup electrodes, filled with Parker Spectra 360 electrode gel, placed 9 cm apart on the forehead. Accuracy and reproducibility of the equipment used to obtain the physiological measures (pen driver, chart speed, and linear response of the polygraph) were within manufacturer’s specifications ( ? 5%) (Grass Instruments, 1986). Values for each physiological variable were measured during the last 2.5 min of each testing occasion, to allow sufficient time for subject acclimatization to the recording situation. An average per 15 s segment was obtained for each measure per subject. These 10 averages for the 2.5 min for each physiological measure (10 X 15 s = 2.5 min) were then summed and divided by 10 to provide the overall average per measure per subject that was used for data analysis. Since the polygraph
382
RESEARCH IN NURSING & HEALTH
presented the levels of each measure in terms of millimeters of pen deflection, these analog values were manually converted to numerical ones using standard Grass conversion factors. Agreement between individuals performing this task and calculating averages was measured at intervals of 2 to 3 weeks and/or whenever a new person started. Percent agreement computed between scorers per physiological measure for the same subject and testing session was maintained at 95% or greater. Nonverbal voice measures were derived from vocalizations audiorecorded with a microphone 15 cm directly in front of the subject’s mouth (for tenseness) and an accelerometer taped to the subject’s neck directly over the thyroid cartilage (for pitch and jitter). A uniform segment from the middle of each vowel phonation was digitized and subsequently processed using computer programs written by one of the authors (YH). Accuracy of the audiotape recorder and signal filtering apparatus was tested and certified by a factory service technician to be within manufacturer’s specifications. The accuracy and reproducibility of the digitizing hardware and software were excellent with an error rate of
Validity and Reliability of Nonverbal Voice Measures as Indicators of Stressor-Provoked Anxiety Barbara F. Fuller, Yoshiyuki Horii, and Douglas A. Conner
To determine the validity and reliability of vocal jitter, tenseness, and pitch as measures of stressor-provoked anxiety, 88 women representing three coping styles (Highly Anxious, Truly Low Anxious, and Repressor) phonated vowels, reported anxiety, and provided measures of muscle tension, heart rate, and sweating 2 weeks before, the day before, and l week after stressful exams. Greater self-reported anxiety, sweating, and heart rate on the day before versus the other occasions confirmed the day before was the most stressful occasion. Results suggest excellent validity and reliability for Jitter as an indicator of stressor-provoked anxiety. @ 1992 John Wiley & Sons, Inc.
Lowery (1987) presents a useful model for nursing research concerning stress. This model encompasses the stressor (life events provoking stress-related responses), mediators (social support, coping behaviors, coping styles), responses (psychological and biological reactions), and outcomes (health, illness), together with their reciprocal influences. A commonly measured psychological reaction is anxiety (Robinson, 1990; Scott, Oberst, & Dropkin, 1980). Biological reactions, termed physiological arousal, include increased blood pressure, heart rate, respirations, sweating, and skeletal muscle tension (Cannon, 1929; Guyton, 1986; Robinson, 1990). The use of these biological reactions to indicate anxiety in nursing assessments is hindered by several limitations. First, measures of physiological arousal vary greatly among individuals (Hugdahl, 1981; Lacey & Lacey, 1958). Second, only weak relationships exist among physiological measures or between groups of physiological measures and groups of psychological measures of anxiety (Morrow & Labrum, 1978; Nesse et al., 1985; Tyrer & Lader, 1976). Third, pathology (Guyton,
1986; McCance & Heuther, 1990) and medications (Swonger & Matejski, 1991), alter physiological measures. Last, methods for obtaining physiological measures can be intrusive. Psychological measures of anxiety also have limitations. The clinical usefulness of self-reported anxiety is hampered by its amenability to intentional or unintentional misrepresentation (Fagin, 1987). Unintentional misrepresentation of self-reported anxiety can stem from certain coping styles (i.e., styles of perceiving, evaluating, and expressing stressor-related anxiety). One coping style, termed Highly Anxious, is a hypervigilance concerning anxiety-related thoughts (Lazarus & Folkman, 1984) that results in high levels of both self-reported anxiety and physiological arousal (Weinberger, 1990). Another coping style is Repression. Repressors exhibit elevated thresholds for recognizing disturbing feelings of anxiety or stress (Krohne & Rogner, 1982) and prefer socially desirable attributes (Bemhardson, 1%7; Byme, 1964). They consistently address anxiety by minimizing cognitions about it and expressions of it (Lazarus & Folkman, 1984). That their level of physiological
Barbara F. Fuller, PhD, RN, is a professor, School of Nursing; Yoshiyuki Horii, PhD, is a professor, Department of Speech Science and Communication Disorders, and Douglas A. Conner, PhD, is a research fellow, School of Nursing; all are at the University of Colorado. This study was supported by Grant NR01468 from the National Center for Nursing Research, National Institutes of Health, to the first two authors. The authors are grateful to Dr. Carol Vojir for statistical advice. This article was received on September 6, 1991, was revised, and accepted for publication April 27, 1992. Requests for reprints can be addressed to Dr. Barbara F. Fuller, Campus Box C-288, University of Colorado Health Sciences Center, 4200 East Ninth Ave., Denver, CO 80262.
8 1992 John Wiley & Sons, Inc. CCC 0160-6891/92/050379-11
380
RESEARCH IN NURSING 8. HEALTH
arousal is similar to, or greater than, levels of Highly Anxious subjects repudiates their social expressions of low anxiety (Asendorpf & Scherer, 1983; Fuller, 1992; Jamner & Schwartz, 1986; Weinberger, 1990). Truly Low Anxious individuals report low levels of anxiety and their physiological arousal is lower than that of Highly Anxious and Repressor subjects (Asendorpf & Scherer, 1983; Weinberger, 1990). The lack of strong correlations between selfreported anxiety and either physiological arousal measures (Nesse et al., 1985) or nursing assessments (Kruger, 1983) may have resulted from combining findings from subjects with different coping styles. A noninvasive, interval-level measure of anxiety would facilitate clinical nursing research with patients who are unable to provide reliable self-reports (e.g., confused elderly). Ideally, such a measure should not differ across coping styles nor be open to intentional misrepresentation. Potential stressor-induced alterations in acoustic voice properties seemed promising as such a measure. Combinations of acoustic properties can differentiate among infants’ “hungry,” “fussy,” and “pain” cries (Fuller, 1991). Vocalizations are composed of combinations of simple sound waves. Each wave has a characteristic frequency and amplitude. Frequency is reported in cycles per second (Hz). Amplitude (the energy contained in the waveform) is measured in decibels (dB). A voice spectrum is a plot of the frequency and amplitude of all the simple waves in a vocalization. The lowest frequency of simple sound waves, termed the fundamental fiequency, is pitch. Variability in the durations of individual sound waves of pitch is termed jitter and is perceived as roughness. Tenseness indicates the relative amount of sound energy in the higher versus lower frequencies in the spectrum (Laver, 1980). Tense voices sound strident or metallic. Sensitive listeners can perceive larger variations in these acoustic properties. Laboratory analysis is required to detect smaller variations. Scherer (1986) theorized that stressor-provoked physiological arousal affects the voice in two ways. First, the arousal increases the tautness of laryngeal and vocal fold muscles which increases pitch and may also increase jitter. Second, the arousal constricts the superior larynx and pharynx and tenses the remaining supraglottal airway muscles which, collectively, elevate the ratio of energy in the higher to lower voice frequencies, thereby increasing tenseness. Accordingly, the major aim of this study was to ascertain the validity of one or more nonverbal voice characteristics (pitch, jitter, and/or tenseness)
as a noninvasive measure of stressor-provoked anxiety. This aim was implemented by (a) determining whether the values for each voice measure differed between an occasion of greater stressorprovoked anxiety (the day before the subject’s graduate comprehensive examination) and two occasions of lesser stressor-provoked anxiety (2 weeks before and 1 week after the exam), and (b) determining if any difference in voice measure values occurred for subjects as a function of coping style: Truly Low Anxious, Highly Anxious, and Repressor. Because the system for classifying coping styles (Weinberger, S c h w a , & Davidson, 1979) and the studies that validated this system used only laboratory stressors, an ancillary study aim was to determine the psychological and biological responses of subjects to a naturally occumng stressor by coping style. The research hypotheses were: 1. Self-reported state anxiety for all subjects will be greater on the day before than on either of the other two less anxiety provoking occasions. 2. Heart rate, palmar sweating, and forehead muscle tension for all subjects will be greater on the day before than on either of the other two occasions. 3. The self-reported anxiety of Repressors will: (a) not differ from, or be lower than, that of Truly Low Anxious subjects; and (b) be lower than that of Highly Anxious subjects. 4. Highly Anxious and Repressor subjects will demonstrate greater physiological arousal than Truly Low Anxious subjects. 5 . The vocal measures of stressor-inducedanxiety will: (a) be greater on the day before than on either of the other two occasions, (b) show variation by stressor condition regardless of coping style, and (c) not differ across coping styles. 6. The short-term (2 min) stability of pitch, jitter, and tenseness measurements will be excellent (r 3 .85).
METHOD
Subjects The convenience sample consisted of 88 healthy female postbaccalaureate health care students from a university in the Westem U.S. who signed an informed consent, had no known pathology, and reported taking no medications except occasional aspirin for headache. Women were used because
NONVERBAL VOICE MEASURES / FULLER ET AL.
(a) the potential subject pool contained few men, and (b) voice data for men and women cannot be combined due to gender differences in pitch. Ages ranged from 22 to 58 years, with a mean of 3 1.9 & 7.4 (SD)years. Thirty-one women were classified as Truly Low Anxious, 31 as Highly Anxious, and 26 as Repressors according to the method described below.
Measures Combinations of the normative median score (13) on the Taylor Manifest Anxiety Scale (MAS) (Taylor, 1953) and the normative upper quartile scores (18 and above) on the Marlowe-Crowne Social Desirability Scale (SDS) (Crowne & Marlowe, 1964) were used, per the method delineated by Weinberger and coworkers (1979), to classify persons as: (a) Truly Low Anxious (MAS score < 14 and SDS score < 18), (b) Highly Anxious (MAS score > 13 and SDS score < 18) or (c) Repressor (MAS score of < 14 and SDS score > 17). Studies concerning construct validity of the Weinberger method in men (Weinberger et al, 1979; Asendorpf & Scherer, 1983) showed that Repressors responded to stressful laboratory tasks with greater palmar sweating and forehead muscle tension, but lower self-reported anxiety, than did Truly Low Anxious subjects. Highly Anxious subjects reported greater state anxiety than did Truly Low Anxious subjects or Repressors. Although one stressful laboratory task did not affect heart rates differently across coping styles in one study (Weinberger et al., 1979), Highly Anxious and Repressor subjects of both sexes responded to another stressful laboratory task with faster heart rates than did Truly Low Anxious subjects (Asendorpf & Scherer, 1983). A similar stressful laboratory task also prompted a greater elevation in the diastolic blood pressure of Repressors than of Truly Low Anxious or Highly Anxious subjects of both sexes (Kiecolt-Glaser& Greenberg, 1983). In a recent study (Fuller, 1992), female Repressors reported less state anxiety and evidenced lower vagal cardiac tone (as measured by respiratory sinus arrhythma and interbeat interval variability) in response to a naturally occumng stressor than either Truly Low Anxious or Highly Anxious females. Truly Low Anxious subjects reported less state anxiety and showed greater vagal cardiac tone than did Highly Anxious subjects. Content validity of the SDS as a measure of affect inhibition and repressive defensiveness (Crowne & Marlowe, 1960, 1964; Millham &
381
Jacobson, 1978) was established by a panel of 10 experts. Correlationsof the SDS with anxiety scores were between -.2 and -.4 (Millimet, 1970). No gender differences were evident in normative scores; stability (> 1 month) was 3 8 , as was internal consistency (Crowne & Marlowe, 1964). The internal consistency of the SDS in this study was 30. Anxiety was measured with the Taylor Manifest Anxiety Scale (Taylor, 1953) and Speilberger’s State Anxiety Inventory (Spielberger, Gorsuch, & Lushene, 1983). Extensive reliability and validity data have been reported for both (Buss, 1955; Hoyt & Magoon, 1954; Spielberger, 1975; Taylor, 1953). No gender difference emerged in normative scores for either tool. Alpha coefficients were 0.85 for the Manifest Anxiety Scale and 0.96 for the State Anxiety Inventory in this study. Stressor-induced physiological arousal was measured with heart rate, palmar sweating, and forehead muscle tension. The validity and reliability of these responses are widely accepted (Cannon, 1929, Guyton, 1986; Morrow & Labrum, 1978; Nesse et al., 1985; Robinson, 1990; Tyrer & Lader, 1976). Physiological measures were simultaneously recorded on moving chart paper using a Grass Model 79D polygraph. For heart rate, Q-waves were obtained from plate electrodes attached on the right leg and forearm. These waves were converted by the tachograph channel into beatshin. Palmar sweating was recorded inversely as Kohms of skin resistance. Resistance was measured using two Ag/AgCI electrodes, filled with isotonic paste, attached to opposite dorsal and ventral surfaces of the middle finger of the nondominant hand. The greater the palmar sweating, the lower the resistance. Forehead skeletal muscle tension, expressed in microvolts, was measured using two stainless steel 9-mm cup electrodes, filled with Parker Spectra 360 electrode gel, placed 9 cm apart on the forehead. Accuracy and reproducibility of the equipment used to obtain the physiological measures (pen driver, chart speed, and linear response of the polygraph) were within manufacturer’s specifications ( ? 5%) (Grass Instruments, 1986). Values for each physiological variable were measured during the last 2.5 min of each testing occasion, to allow sufficient time for subject acclimatization to the recording situation. An average per 15 s segment was obtained for each measure per subject. These 10 averages for the 2.5 min for each physiological measure (10 X 15 s = 2.5 min) were then summed and divided by 10 to provide the overall average per measure per subject that was used for data analysis. Since the polygraph
382
RESEARCH IN NURSING & HEALTH
presented the levels of each measure in terms of millimeters of pen deflection, these analog values were manually converted to numerical ones using standard Grass conversion factors. Agreement between individuals performing this task and calculating averages was measured at intervals of 2 to 3 weeks and/or whenever a new person started. Percent agreement computed between scorers per physiological measure for the same subject and testing session was maintained at 95% or greater. Nonverbal voice measures were derived from vocalizations audiorecorded with a microphone 15 cm directly in front of the subject’s mouth (for tenseness) and an accelerometer taped to the subject’s neck directly over the thyroid cartilage (for pitch and jitter). A uniform segment from the middle of each vowel phonation was digitized and subsequently processed using computer programs written by one of the authors (YH). Accuracy of the audiotape recorder and signal filtering apparatus was tested and certified by a factory service technician to be within manufacturer’s specifications. The accuracy and reproducibility of the digitizing hardware and software were excellent with an error rate of
