Reducing Teacher Stress
Michael R. Bertoch, Elwin C. Nielsen, Jeffrey R. Curley, Walter R. Borg
Utah State University
A prototype treatment developed to significantly reduce symptoms of stress among
inservice teachers was tested in this experiment. Thirty participants selected for high stress levels were randomly assigned to treatment and control groups. They were assessed on environmental, personality, and emotional variables, using self-report and expert-judge measures, at both pre- and posttreatment. The experimental treatment was holistic, incorporating all processes previously found to be related to reducing teacher stress. At posttreatment, the treatment group averaged 1.02 standard deviations lower on the stress measures than the control group. Significant differences in the posttest means, favoring the experimental group, were found for 23 of the 39 variables measured on the three self-report instruments. As a group, the participants demonstrated substantially lower stress levels than the control group after the treatment, with a substantial decrease from their pretreatment stress levels. Since the control group received no treatment, some of the difference may be due to Hawthorne effect.
The Present Study reflects the authors’ concern with the serious threat to teacher mental health caused by occupational stress. Teacher stress is recognized as serious by virtually everyone who has studied the problem (Phillips & Matthew, 1980). A recent search of the ERIC database revealed a substantial amount of descriptive and correlational research regarding teacher stress. However, an extensive literature review failed to produce any reports of projects that used experimental design to evaluate the validity of stress reduction treatments by demonstrating reductions in stress symptomatology. Descriptive and correlational studies have provided important information on possible causal factors. However, these studies are frequently restricted because of research design characteristics and theoretical limitations. The authors’ interest was to develop and evaluate a prototype treatment focused on the apparent causal factors of stress, utilizing the most promising treatment strategies that have emerged from previous research.
The Journal of Experimental Education 57(1), pp. 117-128, Fall 1988.Reprinted with permission of the Helen Dwight Reid Educational Foundation. Published by Heldref Publications, 1319 Eighteenth St., NW, Washington, DC 20036-1802. Copyright © 1988.
Selye’s (1956) establishment of a final common pathway of physiologic responses evoked by stressful events has provided an anchor for a body of theory and research directed toward identifying unique “evocative agents” that result in stress. This work has prompted research along many lines of inquiry. However, a critical mass of knowledge seems to have been achieved in recent years. This knowledge has allowed several integrative paradigms to emerge (Derogatis, 1987; Lazarus, 1966, 1981; Osipow & Spokane 1983, Pettegrew & Wolf, 1982). These paradigms related many of the variables identified by previous research that have stress- inducing potential.
One’s current stress level may be determined by an interaction among the stress events taking place in the environment, the nature and intensity of resulting emotional responses, and personality characteristics of the individual. The exact nature of the relationships among these three areas is incompletely known. Substantial evidence exists, however, to suggest that stressors from these three sources may potentiate one another, and they are at least additive (Derogatis, 1987). It follows that assessment and treatment of persons experiencing high levels of stress may benefit from consideration of these three sources and their possible interactions.
A survey of recent studies of teacher stress shows that many identified stressors appear consistently and may be subsumed under the general domains of environmental and personality- based stressors. Environmental stressors include student discipline and attitude problems, teacher competence, and teacher-administrator relations. Additional stressors include accountability laws, large classes, low salaries, intense pupil dependence, and declining community support. Sources of personality-induced stressors relate to one’s self-perception. Negative self-perception, negative life experiences, low morale, and a struggle to maintain personal values and standards in the classroom all take their toll (Goodman, 1980; Schnacke, 1982; Schwanke, 1981).
Emotional response sets that may contribute to high stress levels have been reported rarely. This may be due to an unrecognized need to identify this class of stressor separately from environmental and personality variables. Negative emotions, however, have been reported as effects of stress. Correlations have been noted between high stress levels and anger, self-doubt, lack of confidence, exhaustion, hypertension, absenteeism, and early retirement.
In summary, the emerging views of the concept of stress and the identification of stressors from many aspects of the teaching profession suggest the futility of trying to remediate teacher stress with a univariate intervention. Stress operates in many dimensions, and it is not always predictable. The authors believe, therefore, that stress management must be conceived and implemented from a holistic perspective with consideration of many research- and theory-based sources of stressors. Similarly, the authors believe that interventions would be enhanced by incorporation of all processes previously found to be effective in reducing teacher stress.
This article represents the development and validation of a treatment program designed to significantly reduce symptoms of stress among inservice teachers whose pretreatment assessment indicated high levels of stress. A prototype treatment was administered to a sample of teachers screened for high stress levels. An experimental design was developed to test the
null hypotheses: (a) there will be no difference between the stress level of teachers who complete the experimental treatment and comparable control group teachers, and (b) there will be no difference between the pre- and posttreatment stress level of teachers who complete the experimental treatment. Hypothesis b is secondary and is concerned with experimental group gains rather than experimental-control differences. We believe that testing this hypothesis adds to the reader’s insight into treatment effects.
Participants The accessible population was public school teachers in northern Utah. Teachers
from middle and high schools gained admission to the program through a three-stage process: (a) submitting an application after a presentation about the program at school faculty meetings, (b) scoring in the top 30 on a screening measure indicating stress level (TSM, see Measures section), and (c) being randomly assigned to treatment (n = 15) or control (n = 15) conditions. Experimental group members completing the treatment received four units of graduate credit in education, and those in the control group were given priority for the treatment during replication. Demographic information is summarized in Table 1.
Treatment Various processes were used in the 12 2-hour treatment sessions, including lecture- discussion, small group sharing of progress and problems, audiovisual presentations, written test evaluations, and homework. Two experienced clinical psychologists conducted the treatment sessions. Activities from the past week(s) were reviewed at the beginning of each session. Session content was as follows1:
Session 1, Introduction. Administrative details, program content, and processes were covered. The clinicians managed the group-forming process while establishing norms of participation, respect, and openness, and modeled the relaxation response.
SUMMARY OF DEMOGRAPHIC INFORMATION
Treatment (n = 15)
Male/Female 6/9 Average age (years) 38.1, SD = 8.28 Married/divorced 15/0
Control (n = 15)
6/9 38.1, SD = 6.99
13/2 3.1 9.1 2.2 73%c 9.5 20% 40% 0% 100%
A verage number of children
A verage years of teaching experience Number of schools of employment Percentage of spouses employed
A verage years spouses employed Pervious mental health care
Alcohol use (moderate) Nonprescription drug use
2.5 9.5 2.1
60% 9.2 27% 27% 0% 100%
Session 2, Concept of Stress. Stress, distress, eustress, Type A and B personality characteristics, and other manifestations of stress were covered. Stages, common causes, consequences, and symptoms of stress were presented.
Session 3, Task-Based and Role-Conflict Stress. Task-based and role related stress were compared. Participants’ unique stressors were identified with force-field analysis planning sheets and stress logs. Group members shared individual analyses in small groups.
Session 4, Assertiveness Life Style. The importance of assertiveness was discussed, along with confusions, myths, differentiation from aggression, and the relationship of assertiveness to self- confidence.
Session 5, Relaxation and Breathing. xperiential breathing and relaxation processes were introduced. A process of systematic relaxation of all muscle groups was then practiced. Members were encouraged to practice regularly until a “relaxation response” became automatic. Session 6, Meditation. Meditation was described as an alternative to achieve a deeper level of relaxation and of contact with the self. All subsequent sessions were initiated with a short session of guided relaxation or meditation.
Session 7, Nutrition. A nutritional evaluation inventory was discussed relative to the participants’ current diets. A lecture-discussion of nutritional habits important in stress management followed, with individual commitment to make changes.
Session 8, Exercise, Mini-Relaxation, and Stretching. A physical exercise evaluation provided individual assessment of needs in this area. Group discussion furthered insight into personal needs for more exercise and methods that could be used. Mini-relaxation and stretching exercises were taught.
Session 9, Holistic Living, Mind and Body. The concept of mindfulness, defined as awareness of self and environment, and awareness of choice and personal creativity were discussed. The importance of making a balance in one’s life was emphasized.
Session 10, Coping with Disappointment and Chemical Stressors. The place of disappointment in the development of stress was discussed. Participants discussed customary ways of coping with disappointment and explored less stressful alternatives.
The endocrine system was described briefly to show how sympathomimetic agents such as caffeine and nicotine trigger an elevated baseline of activity. Agents that reduce this baseline of stress, such as alcohol, minor tranquilizers, barbiturates, and narcotics, were discussed.
Session 11, Support System, Life Stressors, and Teacher Stress. The importance of having an adequate social support system, both at work and in one’s personal life, was discussed.
It was emphasized that to maintain balances, information about stress, and the various coping ideas and techniques need to be utilized from day to day.
Session 12, Understanding Situations, Letting Go of Resentments, and Where to from Here. A review of the experiences and learning from the previous sessions was held, with planning to maintain gains made during the workshop.
Multiple measures of stress were used, as recommended by Bergin and Lambert’s (1978) review of therapeutic outcome research. Ratings were completed by participants, clinicians, and an independent rater. Participants completed a pre- and posttreatment test battery and videotaped clinical interviews. Follow-up data will be collected upon commencement of the replication phase.
Structured Clinical Stress Interview. Recent meta-analytic studies (Edwards, Lambert, Moran, McCulley, Smith, & Ellingson, 1984; Lambert, Hatch, Kingston, & Edwards, 1986) support the authors’ view that interviews by experienced clinicians may assess stressors not probed by self- report measures, thus providing a more complete picture of the subjects’ stress levels. The Structured Clinical Stress Interview (SCSI) was developed to provide a uniform format covering participants’ current or recent stressors, environmental context and possible precipitants, behavioral and physical symptoms, and self-rating of stress level. Interviews were conducted during the week before and after the treatment. The interviewers were blind to group assignment at the pretest but not at the posttest, since the clinicians were involved in the treatment (fiscal restraints prohibited independent interviewers). A third clinician with more than 35 years of experience, blind to groups and sequence, rated a random sample of pre- and posttest interviews taps (n = 16) on the SCSI to provide a reliability check. Videotapes of four pre- and four posttreatment interviews (SCSI) with the two clinicians were collapsed into a single group to provide an adequate number of cases. Correlations with the independent clinician’s ratings yielded an r = .66. This correlation represents a minimum estimate of interrater reliability because the sample of interviews was divided between the two clinicians.
Self-Report Measures. In addition, stress level was assessed across 39 variables using three self- report measures that completed the assessment battery: the Derogatis Stress Profile (DSP; Derogatis, 1987), the Occupational Stress Inventory (OSI; Osipow & Spokane, 1983), and the Teacher Stress Measure (TSM; Pettegrew & Wolf, 1982).
The DSP consists of 77 items that assess stress levels in environmental personality, and emotional domains. The developers reported alpha reliabilities on 11 subscales ranging from .79 to .99. Test-retest coefficients ranged from .79 to .93 on the subjects and was .90 for total scores (see Table 4). Some evidence of construct validity and predictive validity was also reporter.
The OSI consists of 140 items that assess stress levels in three dimensions of occupational adjustment: occupational stress, psychological strain, at coping resources. The developers reported alpha reliabilities on the three dimensions of .89, .94, and .99, respectively, and on the 14 subscales from .71 to 94. Two-week test-retest coefficients from .88 to .94 were reported on the three dimensions, and from .56 to .94 on the individual scales. Some evidence of construct and concurrent validity was also reported.
The TSM consists of 70 items that assess stress levels on 14 variables. The author reported alpha reliability coefficients ranging from .57 to .91. Median reliability was .82 with only two scales below .75. Some evidence, of concurrent validity was also reported.
The above measures represent the most promising measures identified in extensive review of contemporary instruments. Each self-report measure and the SCSI yielded total scores based on a 5-point (SCSI, DSP, OSI) or 6-point (TSM) scale. These provided the main indices of change in teacher stress. To estimate the concurrent validity for the self-report measures using the clinical interview as a criterion, and to provide a rationale for including all measures in an assessment battery, correlations were computed between pretest scores across groups (N = 30). Ninety-five percent confidence intervals (95% CI) were determined based on Fisher’s Z transformations used to provide a normal distribution. The results presented in Table 2 show correlations ranging from .56 to .72, suggesting that the measures are moderately correlated but do not all measure the same construct. Correlations between the SCSI and the DSP, OSI, and TSM were moderate and consistent at .58, .56, and .57, respectively.
PEARSON CORRELA TIONS BETWEEN MEASURES ON PRETESTS (N = 30)
Measure r DSP .58
OSI .56 TSM .57
< .001 < .001 < .001
.280-.884 .245-.762 .260-.770
.56 < .001 .58 < .001
.72 < .001
The first hypothesis, which predicted no differences between the stress levels of participants who completed the treatment and a comparable control group, was examined using analyses of covariance (ANCOVA) between-group posttest scores from the DSP, OSI, TSM, and SCSI, with the pretest scores entered as covariates. This analysis allowed the slope relating the pretest and posttest to be estimated rather than forced to be 1, as when gain scores alone are used as the dependent variable, thus providing a more sensitive test due to reduced error variance (Hendrix, Carter, & Hintze, 1978; Linn & Slinde, 1977).
The second hypothesis, which produced no difference between the pre- and posttest stress level of participants who received the treatment, was examined using correlated means t tests. Since the 12 treatment sessions concluded near the end of the school year—a period described by teachers as highly stressful—this analysis provided information on the direction of change (i.e., whether the treatment group improved or the control group deteriorated; see Table 3).
To determine whether the measures employed provided comparable data for assessing treatment effects, an effect size (ES) was computed for each measure. Finding were thus transformed into a common metric (standard deviation units), rendering an index of the magnitude of effect or change.
To provide information concerning differences between the experimental and control groups on the subscores obtained from the self-report measures, ANCOVAs were computed between
groups on posttreatment means scores with pretreatment scores entered as covariates. This analysis, which indicated variables showing significant change at the posttest, will be used, along with data from participant feedback forms completed after each session, to suggest possible refinements in the treatment program and the instrumentation (see Table 4). However, because of the small number of cases and the low reliability of some of the subscores, data in Table 4 should be regarded as tentative.
After the treatment, the experimental group demonstrated substantially lower stress levels than control group members. Significant differences between experimental and control groups in adjusted means were found on the OSI, DSP, TSM, and SCSI (see Table 3). Computation of effect sizes for the above measures indicates how many standard deviations the treatment group differed from the control group at posttreatment. A substantially lower stress level, averaging 1.02 SD, was found to be associated with participation in the treatment.
The experimental group demonstrated a substantial decrease in their stress level after the treatment. Table 3 shows that DSP means decrease from 152.60 pretreatment to 117.30 posttreatment. Similar decreases in means were observed on the OSI, TSM, and SCSI, with all four ps significant at the .001 level.
Control group means on the DSP decreased from 142.40 to 132.87 pre- to posttreatment, an improvement significant at the .05 level. On the SCSI, however, a significantly higher stress level was indicated at posttreatment (.05 level). The OSI and TSM showed very small changes that were not statistically significant.
Although the experimental group indicated higher stress levels than control group on all pretreatment measures (p = .16, .17, .024, and .021 on the DSP, OSI, TSM, and SCSI, respectively), they were significantly on all posttreatment means adjusted for pretreatment scores. A concern regarding internal validity in studies using samples selected for extreme scores is that statistical regression to the mean may account for treatment gains (Borg, 1987; Borg & Gall, 1983; Kazdin 1980). Since the subjects with the highest scores were randomly assigned to treatment or control, both groups should have regressed a like amount. Had regression accounted for a significant increment of stress reduction in the treatment group, a similar change in the control group would be expected. However, although the treatment group’s average means across measures dropped from 210.30 pretreatment to 178.55 posttreatment, corresponding control group means dropped only 1.3 points, from 193.38 to 192.08 during the same period. The relative stability of the control group average scores suggests that regression to the mean may have occurred in the context of a more stressful posttreatment environment, thus not detectable in score changes. However, large differences in experimental group scores also suggests that gain associated with the treatment was not confounded with regression effects.
SUMMARY OF t TESTS, ANCOVAS, AND EFFECT SIZES
Mpre, SD Mpost, SD
Mpre, SD Mpost, SD
F(1, 27) p Mc,adja, Mt,adja, SDpoola
142.40, 24.01 132.87,25.56
152.60, 24.08 117.30, 43.00
t Tests: Control group pre-post
232.40, 34.98 224.90, 43.60
42.20, 6.98 48.47, 9.56
50.60, 11.16 40.00, 10.38
50.25 38.22 9.86
t Tests: Treatment Group pre-post
ANCOVAs between treatment and control posttests
6.02 8.35 .021 .008
137.68 368.05 112.44 342.22 25.24 34.37
238.18 195.49 38.79
Mc,adja, and Mt,adja, = posttest means adjusted for the pretest means (i.e., the covariate), SDpool = pooled SDs from
control pre- and posttests and treatment pretests. On all measures higher scores indicated higher stress. All t tests were two-tailed.
The present study clearly demonstrates a reduction in teacher stress by subjects in the experimental treatment. Table 4 shows that significant experimental vs. control differences were found on 23 of the 39 variables measured by the self-report instruments, many of which may be related to specific treatment content. Interestingly, treated participants scored significantly higher (i.e., less favorably) on the self-care and rational/cognitive coping subscales of the OSI after the treatment. This finding correlates with comments from participants on feedback forms from later sessions such as, “It all seems so helpful, but I need more practice on the things learned,” “We covered it all—very quickly,” or even “A little bit of ignorance is bliss.” Taken together, these may suggest that a sense of overload and need for integration may have been caused by the numerous methods employed reduce stress, paradoxically introducing a new source of stress. In response to a question on the feedback questionnaires asking which areas had most beneficial, individual participants stressed different areas. This further suggests that increased reduction of stress in individual cases may be enhanced through idiographic pretreatment assessment and a more focused treatment based on individual needs. An understanding of the process variables imbedded in the treatment package will require further research to determine whether specific factors emerge, thus allowing a better match to individual needs.
VARIABLES AND SIGNIFICANCE OF CHANGES BETWEEN
PRE- AND POSTTREATMENT FOR TREATMENT VS. CONTROL GROUP
Subscale Alphaa DSP
Time Pressure .93 Driven Behavior .88 Attitude Posture .86 Relaxation Potential .91 Role Definition .90 Vocational Satisfaction .79 Domestic Satisfaction .86 Health Posture .85 Hostility .81 Anxiety .84 Depression .85 OSI
Role Overload .83 Role Insufficiency .90 Role Ambiguity .78 Role Boundary .82 Responsibility .71 Physical Environment .85 Vocational Strain .71 Psychological Strain .89 Interpersonal Strain .81 Physical Strain .87 Recreation .71 Self-Care .73 Social Support .83 Rational/Cog. Coping .78 TSM
Role Ambiguity .79 Rover Overload .76 Role Conflict .82 Nonparticipation .76 Role Preparedness .57 School Stress .89 Job Satisfaction .86 Management Style .74 Life Satisfaction .91 Task Stress .84 Supervisory Support .89 Peer Support .84 Untitled NA Illness Symptoms .82
a Alpha reliability coefficients * p<.01. **p<.05.
Pre Post Pre Post
18.2 15.8 17.1 15.5 12.9 10.3 12.2 11.7 16.1 13.5 15.5 14.5 15.5 10.6 13.2 12.2 12.5 8.7 11.7 10.6 13.3 11.7 12.5 11.7 12.5 8.8 11.1 10.4 11.7 9.1 11.3 10.2 10.6 8.7 10.7 10.9 18.3 12.6 16.3 14.8 11.1 7.6 10.9 10.4
32.3 29.5 30.3 29.1 28.7 23.9 28.3 28.3 26.5 22.3 23.3 23.5 26.3 20.1 25.7 23.0 29.4 24.1 28.3 26.9 15.7 14.5 14.5 14.9 24.5 18.4 21.9 22.1 30.2 21.3 24.7 23.3 27.8 21.2 26.5 22.9 28.0 21.9 22.6 22.9 22.0 25.0 24.5 26.3 22.5 31.3 23.7 27.4 24.7 39.7 36.7 40.1 28.8 35.2 29.4 31.2
15.3 12.1 12.8 11.6 21.0 16.9 18.0 17.9 20.7 17.3 18.5 17.7 19.1 13.5 17.1 16.3 15.1 12.9 15.0 14.6 18.5 16.5 15.8 16.5 18.1 15.1 16.5 15.1 16.7 12.5 15.7 15.1 16.3 11.9 13.3 12.9 40.5 34.9 35.9 38.2 12.1 9.7 9.7 8.7
9.1 6.5 7.8 7.9 21.9 17.2 20.8 19.8 16.2 11.8 15.5 12.7
as reported in the test manuals. ***p<.01.
0.447 2.975* 2.977* 3.061* 4.956** 0.293 3.570* 0.139 4.596** 3.667** 5.237**
0.748 5.123** 0.107 6.666** 1.981 1.666
17.532*** 4.910** 4.124* 6.159**
–0.392 –4.807** –0.582 –7.410**
0.199 4.312** 4.253 4.767** 7.173** 6.175** 0.300
13.040*** 5.652** 9.462*** 0.593 4.703*** 3.893* 0.451
The present study addressed a need for remediation of teacher stress with a complex, multifaceted treatment package. As many strategies as practicable were included; some, perhaps, were unnecessary or differentially effective in individual cases. With the main questions resolved—validation the treatment package alters stress levels overall in the desired direction—interest shifts toward more specific concerns. Which of the variables covered in the treatment contribute most to stress reduction? Will a dismantling of components of the treatment package into multiple treatments in understanding the sufficient and necessary conditions of stress reduction? Ultimately, further research is needed to develop a treatment strategy that will vary specific aspects of the treatment with respect to subject variables to determine how to maximize stress reduction within teacher populations (Kazdin, 1980). More immediately, several analyses pertinent to the present study await additional funding and time to complete. Follow-up assessment to determine the durability of the reductions in stress over time are anticipated at approximately 6 months and 1 year, contingent upon grant funding for two replication phases allowing further investigation of important area.
Persons wanting a detailed set of lesson plans for the 12 sessions should contact Professor Michael Bertoch, Department of Psychology, Utah State University, Logan, UT 84322-2810.
Bergin, A. E., & Lambert, M. J. (1978). The evaluation of therapeutic outcomes. In A. E. Bergin & S. L. Garfield (Eds.), Handbook of psychotherapy and behavior changes: An empirical analysis. New York: Wiley.
Borg, W. R. (1987). Applying educational research: A practical guide for teachers. New York: Longman.
Borg, W. R., & Gall, M. D. (1983). Educational Research: An introduction. New York: Longman.
Derogatis, L. R. (1987). The Derogatis Stress Profile (DSP): Quantification of psychological stress. Advances in Psychosomatic Medicine, 17, 30-54.
Edwards, B. C., Lambert, M. J., Moran, P. W., McCully, T., Smith, K. C., & Ellingson, A. G. (1984). A meta-analytic comparison of the Beck Depression Inventory and the Hamilton Rating Scale for Depression as measures of treatment outcome. British Journal of Clinical Psychology, 23, 93-99.
Goodman, V. B. (1990). Urban teacher stress: A critical literature review. ERIC Document Reproduction Service No. ED 221 611.
Hendrix, J. L., Carter, M. W., & Hintze, L. J. (1978). A comparison of five statistical methods for analyzing pretest-posttest designs. Journal of Experimental Education, 47, 96-102.
Kazdin, A. E. (1980). Research design in clinical psychology. New York: Harper & Row.
Lambert, M. J., Hatch, D. R., Kingston, M. D., & Edwards, B. C. (1986). Zung, Beck, and Hamilton rating scales as measures of treatment outcome: A meta-analytic comparison. Journal of Consulting and Clinical Psychology, 54, 54-59.
Lazarus, R. W. (1966). Psychological stress and the coping process. New York: McGraw-Hill Lazarus, R. W. (1981). The stress and coping paradigm. In C. Eisdorfer, D. Cohen, A.
Kleinman, & P. Maxim (Eds.), Models for clinical psychotherapy. New York: Spectrum. Linn, R. L., & Slinde, J. A. (1977). The determination of the significance of change between
pre- and posttesting periods. Review of Educational Research, 47(l), 121-150.
Osipow, S. H., & Spokane, A. R. (1983). Manual for measures of occupational stress, strain
and coping (Form E-2) Columbus, OH: Marathon Consulting and Press.
Pettegrew, L. S., & Wolf, G. E. (1982). Validating measures of teacher stress. American
Educational Research Journal, 19(3), 373-396.
Phillips, B. N., & Matthew, L. (1980). The changing role of the American teacher: Current and future sources of stress. In C. L. Cooper & J. Marshal (Eds.), White collar and professional stress (pp. 93-111). New York: Wiley.
Schnacke, S. B. (1982). Burnout: Coping with predictable professional life crises. Paper presented at the annual meeting of the American Association of Colleges for Teacher Education, Houston, TX. (ERIC Document Reproduction Service No. ED 257 836)
Schwanke, D. C. (1981). Teacher stress: Selected ERIC resources. Washington, DC: ERIC Clearinghouse on Teacher Education. (ERIC Document Reproduction Service No. ED 204– 258)
Selye, H. (1956). The stress of life. New York: McGraw-Hill.