Biofeedback and Self-Regulation, VoL 2, No, 2, 1977
Biofeedback Heart Rate Training during Exercise David S. Goldstein, Richard S. Ross, and Joseph V. Brady The Johns Hopkins UniversitySchool of Medicine
Eighteen healthy human subjects participated in weekly sessions of five lO-minute trials o f walking on a treadmill at 2.5 mph and 6°70 grade. Eight experimental subjects received beat-to-beat heart rate biofeedback during the exercise and were instructed to try to lower their heart rates; ten control subjects did not receive feedback. By the end o f 5 weeks (25 trials), the experimental group showed a significantly lower mean heart rate (96.8 vs. 108.6 bpm), systolic blood pressure (114.0 vs. 131.3 mmHg), and ratepressure product (11.0 X 103 vs. 14.3 X 103 bpm-mmHg) during exercise than the control group. These differences were maintained after crossover o f the feedback provision f o r five more weeks.
Recent research in instrumental cardiovascular conditioning has concerned potential clinical applications of biofeedback in patients with various circulatory disorders. Benson, Shapiro, Tursky, and Schwartz (1971), Patel (1973), Haynes (1972), Miller, DiCara, Solomon, Weiss, and Dworkin (1970), Elder, Ruiz, Deabler, and Dillenkoffer (1972), and Krisst and Engel (1975) have reported some success in instrumentally conditioned decreases in blood pressure in patients with essential hypertension. Bleecker and Engel (1973a,b) reported instrumentally conditioned heart rate changes in patients with atrial fibrillation and instrumental control of the frequency of Wolff-Parkinson-White beats. Engel and Melmon (1971), Weiss and Engel (1971), and Picketing and Gorman (1975) reported instrumentally learned heart rate changes in patients with arrhythmias. So far no one has reported attempts to influence heart rate or blood pressure responses to exercise using biofeedback. If this could be accomplished, the technique might be useful in the treatment of angina pectoris, because: (1) the product of heart rate and systolic blood pressurewthe ratepressure product--is a reasonable indirect measure of myocardial oxygen consumption (Monroe, 1964; Robinson, 1967; Redwood, Rosing, Goldstein, Beiser, & Epstein, 1971; Amsterdam, Hughes, DeMaria, Zelis, & 107 This journal is c o p y r i g h t e d by Plenum. Each article is available f o r $7.50 f r o m PIenurn Publishing C o r p o r a t i o n , 2 2 7 West 17th Street, New Y o r k , N.Y. 10011.
108
Goldstein et al.
Mason, 1974; Bruce, 1974); (2) anginal pain results from an excess of myocardial oxygen consumption over supply (Friesinger and Ross, 1972); and (3) most patients with angina pectoris suffer their chest discomfort during exercise, after their rate-pressure products exceed threshold values (Cohen, Elliot, Rolett, & Gorlin, 1965; Littler, Honour, Sleight, & Stott, 1973). We report in this paper a study of instrumental heart rate training in healthy subjects during mild exercise.
METHODS In order to find out if heart rate biofeedback influences the cardiovascular response to exercise, we conducted a study in which healthy subjects either received heart rate feedback or did not receive feedback during weekly sessions of mild treadmill exercise. We compared the feedback group's mean heart rate, systolic blood pressure, and rate-pressure product with that of the no-feedback group before, during, and after the sessions. We then crossed over the feedback contingency to find out if the feedback group could maintain their performance during subsequent sessions without feedback and to find out if the no-feedback group could change their performance once they received feedback.
Subjects Eighteen persons ranging in age from 18 to 39 years old, including 10 males and 9 females, responded to an advertisement posted in various areas of the Johns Hopkins University and Hospital asking for subjects interested in learning to control their hearts using biofeedback. These 18 subjects constituted the total group of responders to the advertisement who participated in the entire study, and they received $50 each for their participation, independent of their performance. All but one of the subjects reported no history of any medical disease. One subject reported a history of a kidney ailment which had been treated and was asymptomatic at the time of this study. The performance of this subject differed in no way from that of the other members of his experimental group, Group C - E (see below for description of group assignments). None of the subjects showed irregularities of cardiac rate or rhythm during the study, and no subject reported tiring or shortness of breath due to the exercise. The subjects did not take any medication during the study, and, except for an occasional cold, no experimental session was cancelled for health reasons. We therefore considered all the subjects to be in good health during the study.
Biofeedback Heart Rate Training during Exercise
109
Sessions
Each subject participated in 10 weekly sessions in the EKG Stress Laboratory of the Heart Station at the Johns Hopkins Hospital. There were no practice sessions, and the sessions occurred at various times during the day or evening as personal and laboratory schedules permitted. Most of the experimental sessions took place between 5 and 11 PM weekdays and between 9:30 AM and 1 PM weekends. The person conducting subject scheduling did not know the particular group assignments, so that no systematic differences occurred in the timing of experimental sessions among the various groups with respect to either time of day or day of the week. All
SIGNAL LIGHT
~
¢ ~
IfYWZ '.
/ '/
#
\~ t
.+cT~omc DIGITAL
N
CARDIoTACHOM| HANTN EIIc | L~ ' = ~ t
~i
Fig. 1. Overhead schematic view of the experimental setup.
110
Goidstein et al. ttt MINUTES I ........ o
I ........ IO
i20 REST
-
I ...... 30 REST
~ .... 40
I ....... so REST
n......... 60
J ....... 70
q 80
REST
Fig. 2. Sequence of events in a typical session.
the subjects were tested in the same room over approximately the same 3-month time period. Figure 1 shows an overhead schematic view of the experimental setup. In a typical session, upon entry into the laboratory the subject was met by the experimenter (D.G.), who briefly reviewed the subject's performance of previous sessions, inquired about the subject's health and activities that day, and began setting up the experiment. Setting up (Fig. 2) for each session lasted approximately 4 minutes and consisted of: (1) preparing the subject's skin by rubbing with an alcohol dab at the second right intercostal space in the midclavicular line and at the sixth left intercostal space in the anterior axillary line; (2) placing disposable EKG electrodes at these sites; (3) connecting these electrodes to polygraph cables taped to the arm bar of the treadmill; (4) testing the EKG signal during a 10-second walk on the treadmill; (5) calibrating the polygraph; and (6) taping a stethoscope head in the subject's right antecubital fossa, wrapping a blood pressure cuff around it, and inflating the cuff once to verify adequate Korotkoff sounds. After setting up, the subjects stood quietly on "the treadmill for approximately 2 minutes (Fig. 2). Then a 2-minute baseline period occurred for recording baseline heart rate and setting the criterion heart rate--that is, the heart rate below which the subject would be rewarded. The first trial began immediately after the baseline period. Then followed five 10-minute trials of walking on the treadmill at a fixed rate of 2.5 mph and 6°70 grade (Fig. 2). The trials were separated by 4-minute rest periods (Fig. 2), during which the subjects remained standing but could talk, stretch, or sip water. At 500 seconds into each trial, the observer measured the subject's systolic pressure manually using the cuff and stethoscope. After each trial, the mean trial heart rate was calculated by dividing the total number of beats during the trial by 10, and the rate-pressure product was calculated by multiplying the mean trial heart rate by the systolic pressure at 500 seconds. This method of calculating rate-pressure product was used because repeated measurements of blood pressure using the cuff and stethoscope may have interfered with the subject's performance, and single determinations of heart rate would have ignored much of the trial data while increasing the variability of measurement.
Biofeedback Heart Rate Training during Exercise
111
Subjects were admonished from talking during the trials. They could breathe in any manner they wished, tense or relax muscles, or think any thought they pleased, but they were instructed not to close their eyes or to move in such a way that the EKG signal was interfered with. Postures obviously taken to minimize the amount of exercise, such as leaning over the rails of the treadmill and dragging the feet, were neither allowed nor observed. During the trials, the observer sat behind and to the right of the subject out of the subject's view (Fig. 1) and remained silent. Between trials, if asked, the observer informed the subjects about techniques other subjects had used and about possible clinical applications of the results of the study. Feedback Subjects were assigned randomly to either Group E - C (experimental -control) or Group C - E (control-experimental). Subjects in Group E - C received heart rate feedback during the first five sessions of the study but not during the second five sessions, i.e., they began the experiment in an experimental, feedback condition and ended in a control, no-feedback condition. Subjects in Group C - E received no feedback during the first five sessions but did receive feedback during the second five sessions, i.e., they began the experiment in a control condition and ended in an experimental condition. The eight subjects assigned to Group E - C were instructed to try to lower their heart rates during the trials. Of the 10 subjects assigned to Group C - E , 5 (Subgroup T) were instructed for the first five sessions to try to lower their heart rates and 5 (Subgroup N) were instructed to walk normally. For the second five sessions, all the subjects in Group C - E were instructed to try to lower their heart rates. The average ages of the subjects in the various groups were comparable (27 years old for Group E - C ; 26 for Group C - E ; 26 for Subgroup T; and 25 for Subgroup N), as was the sexual composition of the groups (5 males and 3 females for Group E - C ; 5 males and 5 females for Group C - E ; 3 males and 2 females for Subgroup T; and 2 males and 3 females for Subgroup N). Heart rate feedback was provided in four forms for each subject in an experimental condition, as shown schematically in Fig. 1. First, an amber light was illuminated at the end of each R - R interval below a preset criterion. The amber light remained illuminated through the ensuing R - R interval. No light signal followed R - R intervals which were above criterion. Illumination of the signal light, therefore, indicated success at control of heart rate. Second, an electronic display of beat-to-beat heart rate provided additional information beyond the binary feedback of the signal light,
112
Goldstein et al.
because it indicated the degree of deviation from the criterion rate. Third, the polygraph's cardiotachometer channel was pointed out to the subjects and positioned to be clearly visible to them (Fig. 1); this provided information about trends in heart rate over time. Fourth, the observer commented on the subject's performance after--but not duringmeach trial. The subjects chose which feedback form to attend to and could switch from one to another at any time. The general rationale for setting the criterion heart rate was that of shaping, i.e., reinforcing slight changes in performance in the appropriate direction at first and then gradually increasing the difficulty of the criterion as performance improved. The observer set the criterion manually after each trial so that the signal light lit for approximately 80°70 of the beats during the next trial. For the first trial of the day, criteria were adjusted at 1 minute into the trial in order to achieve the 80070 goal. In practice, setting the criterion heart rate at 1-2 bpm above the mean heart rate on the previous trial usually satisfied the 80°70 goal, as did setting the criterion heart rate for Trial 1 of each session at the mean heart rate for Trial 1 of the previous session. Subjects were informed each time a criterion change was made. No subject received feedback during the first trial of the first experimental session in order to obtain preexperimental baseline information on heart rate during exercise for all groups prior to feedback. The feedback provided in these various forms constituted the only reinforcement for performance in the study.
Variables and Data Analysis Independent variables in this experiment were: (1) feedback condition; (2) instructions (i.e., try to lower heart rate vs. to walk normally); (3) sessions; and (4) trials within sessions. Dependent variables included: (I) mean trial heart rate; (2) systolic blood pressure; (3) rate-pressure product, expressed as mmHg X bpm X 10~; (4) baseline mean heart rate; and (5) percent change in heart rate from the baseline to the rate for a given trial during the session. The experimental data were analyzed in two steps. First, analyses of variance (ANOVAs) were performed and then independent means t-tests comparing the experimental groups. The reason for this apprach was that the data analysis involved multiple between-group comparisons, and the probability that at least one of these would reveal a statistically significant difference by chance alone was unacceptably high. ANOVAs provided an assessment of the significance of the overall effects of feedback, sessions, etc., on the dependent variables, and obtaining significant effects in the ANOVAs permitted application of multiple t-tests to the data without fear of wrongly inferring significant differences between groups.
Biofeedback Heart Rate Training during Exercise
113
The type of ANOVA used was a repeated measures design, betweenwithin analysis of variance, with measures taken in the same subjects repeatedly (e.g., over sessions and trials), some independent variables applied within subjects (e.g., sessions and trials), and some others applied between subjects (e.g., feedback). The ANOVAs also permitted assessment of the significance of interaction effects in the data. For instance, if the "Feedback X Trials" interaction effect was significant, this would imply that the effect of trials on the dependent variable was significantly influenced by the feedback provision to the group under consideration. Although ANOVAs allow a number of general inferences from data, they do not pinpoint where the significant effects are. Independent means ,t-tests were, therefore, conducted to compare the experimental groups for each trial and session once significant ANOVAs had been obtained. Onetailed t-tests were used in calculating significance levels, because we wished to test a specifically unidirectional hypothesis, namely, that heart rate biofeedback produces decreases in mean heart rate, in systolic pressure, and in the rate-pressure product, an index of myocardial oxygen consumption. Throughout the study, p values of less than 0.050 defined statistical significance. RESULTS
Sessions 1 through 5 Baseline Heart Rate. Resting (i.e., no exercise) baseline mean heart rates did not differ significantly between Groups E - C and C - E for any of Sessions 1 through 5 (Table I). The resting baseline mean heart rate for Table L Baseline Heart Rate Data for Groups E - C and C - E Group E - C
*
Group C - E
Session
Mean HR
SD a
Mean HR
SD a
df
t
p
1 2 3 4 5
90.0 90.5 89.1 88.5 89,8
15.7 8.0 12.0 9.2 14.9
90.8 84.7 89.9 86.7 85.7
13.9 11.1 10.8 12.3 10.7
16 16 16 16 16
0.11 1.20 0.14 0.34 0.67
n.s. n.s. n.s. n.s. n.s.
6 7 8 9 10
94,4 83,6 84,5 83.9 84.5
13.2 10.0 13.9 14.6 11.4
95.2 92.5 93.1 89.8 93.5
8.2 8.8 11o9 7.1 10.6
16 16 16 16 16
0A6 2.00 1,46 1.13 1.73
n.s.
Biofeedback Heart Rate Training during Exercise David S. Goldstein, Richard S. Ross, and Joseph V. Brady The Johns Hopkins UniversitySchool of Medicine
Eighteen healthy human subjects participated in weekly sessions of five lO-minute trials o f walking on a treadmill at 2.5 mph and 6°70 grade. Eight experimental subjects received beat-to-beat heart rate biofeedback during the exercise and were instructed to try to lower their heart rates; ten control subjects did not receive feedback. By the end o f 5 weeks (25 trials), the experimental group showed a significantly lower mean heart rate (96.8 vs. 108.6 bpm), systolic blood pressure (114.0 vs. 131.3 mmHg), and ratepressure product (11.0 X 103 vs. 14.3 X 103 bpm-mmHg) during exercise than the control group. These differences were maintained after crossover o f the feedback provision f o r five more weeks.
Recent research in instrumental cardiovascular conditioning has concerned potential clinical applications of biofeedback in patients with various circulatory disorders. Benson, Shapiro, Tursky, and Schwartz (1971), Patel (1973), Haynes (1972), Miller, DiCara, Solomon, Weiss, and Dworkin (1970), Elder, Ruiz, Deabler, and Dillenkoffer (1972), and Krisst and Engel (1975) have reported some success in instrumentally conditioned decreases in blood pressure in patients with essential hypertension. Bleecker and Engel (1973a,b) reported instrumentally conditioned heart rate changes in patients with atrial fibrillation and instrumental control of the frequency of Wolff-Parkinson-White beats. Engel and Melmon (1971), Weiss and Engel (1971), and Picketing and Gorman (1975) reported instrumentally learned heart rate changes in patients with arrhythmias. So far no one has reported attempts to influence heart rate or blood pressure responses to exercise using biofeedback. If this could be accomplished, the technique might be useful in the treatment of angina pectoris, because: (1) the product of heart rate and systolic blood pressurewthe ratepressure product--is a reasonable indirect measure of myocardial oxygen consumption (Monroe, 1964; Robinson, 1967; Redwood, Rosing, Goldstein, Beiser, & Epstein, 1971; Amsterdam, Hughes, DeMaria, Zelis, & 107 This journal is c o p y r i g h t e d by Plenum. Each article is available f o r $7.50 f r o m PIenurn Publishing C o r p o r a t i o n , 2 2 7 West 17th Street, New Y o r k , N.Y. 10011.
108
Goldstein et al.
Mason, 1974; Bruce, 1974); (2) anginal pain results from an excess of myocardial oxygen consumption over supply (Friesinger and Ross, 1972); and (3) most patients with angina pectoris suffer their chest discomfort during exercise, after their rate-pressure products exceed threshold values (Cohen, Elliot, Rolett, & Gorlin, 1965; Littler, Honour, Sleight, & Stott, 1973). We report in this paper a study of instrumental heart rate training in healthy subjects during mild exercise.
METHODS In order to find out if heart rate biofeedback influences the cardiovascular response to exercise, we conducted a study in which healthy subjects either received heart rate feedback or did not receive feedback during weekly sessions of mild treadmill exercise. We compared the feedback group's mean heart rate, systolic blood pressure, and rate-pressure product with that of the no-feedback group before, during, and after the sessions. We then crossed over the feedback contingency to find out if the feedback group could maintain their performance during subsequent sessions without feedback and to find out if the no-feedback group could change their performance once they received feedback.
Subjects Eighteen persons ranging in age from 18 to 39 years old, including 10 males and 9 females, responded to an advertisement posted in various areas of the Johns Hopkins University and Hospital asking for subjects interested in learning to control their hearts using biofeedback. These 18 subjects constituted the total group of responders to the advertisement who participated in the entire study, and they received $50 each for their participation, independent of their performance. All but one of the subjects reported no history of any medical disease. One subject reported a history of a kidney ailment which had been treated and was asymptomatic at the time of this study. The performance of this subject differed in no way from that of the other members of his experimental group, Group C - E (see below for description of group assignments). None of the subjects showed irregularities of cardiac rate or rhythm during the study, and no subject reported tiring or shortness of breath due to the exercise. The subjects did not take any medication during the study, and, except for an occasional cold, no experimental session was cancelled for health reasons. We therefore considered all the subjects to be in good health during the study.
Biofeedback Heart Rate Training during Exercise
109
Sessions
Each subject participated in 10 weekly sessions in the EKG Stress Laboratory of the Heart Station at the Johns Hopkins Hospital. There were no practice sessions, and the sessions occurred at various times during the day or evening as personal and laboratory schedules permitted. Most of the experimental sessions took place between 5 and 11 PM weekdays and between 9:30 AM and 1 PM weekends. The person conducting subject scheduling did not know the particular group assignments, so that no systematic differences occurred in the timing of experimental sessions among the various groups with respect to either time of day or day of the week. All
SIGNAL LIGHT
~
¢ ~
IfYWZ '.
/ '/
#
\~ t
.+cT~omc DIGITAL
N
CARDIoTACHOM| HANTN EIIc | L~ ' = ~ t
~i
Fig. 1. Overhead schematic view of the experimental setup.
110
Goidstein et al. ttt MINUTES I ........ o
I ........ IO
i20 REST
-
I ...... 30 REST
~ .... 40
I ....... so REST
n......... 60
J ....... 70
q 80
REST
Fig. 2. Sequence of events in a typical session.
the subjects were tested in the same room over approximately the same 3-month time period. Figure 1 shows an overhead schematic view of the experimental setup. In a typical session, upon entry into the laboratory the subject was met by the experimenter (D.G.), who briefly reviewed the subject's performance of previous sessions, inquired about the subject's health and activities that day, and began setting up the experiment. Setting up (Fig. 2) for each session lasted approximately 4 minutes and consisted of: (1) preparing the subject's skin by rubbing with an alcohol dab at the second right intercostal space in the midclavicular line and at the sixth left intercostal space in the anterior axillary line; (2) placing disposable EKG electrodes at these sites; (3) connecting these electrodes to polygraph cables taped to the arm bar of the treadmill; (4) testing the EKG signal during a 10-second walk on the treadmill; (5) calibrating the polygraph; and (6) taping a stethoscope head in the subject's right antecubital fossa, wrapping a blood pressure cuff around it, and inflating the cuff once to verify adequate Korotkoff sounds. After setting up, the subjects stood quietly on "the treadmill for approximately 2 minutes (Fig. 2). Then a 2-minute baseline period occurred for recording baseline heart rate and setting the criterion heart rate--that is, the heart rate below which the subject would be rewarded. The first trial began immediately after the baseline period. Then followed five 10-minute trials of walking on the treadmill at a fixed rate of 2.5 mph and 6°70 grade (Fig. 2). The trials were separated by 4-minute rest periods (Fig. 2), during which the subjects remained standing but could talk, stretch, or sip water. At 500 seconds into each trial, the observer measured the subject's systolic pressure manually using the cuff and stethoscope. After each trial, the mean trial heart rate was calculated by dividing the total number of beats during the trial by 10, and the rate-pressure product was calculated by multiplying the mean trial heart rate by the systolic pressure at 500 seconds. This method of calculating rate-pressure product was used because repeated measurements of blood pressure using the cuff and stethoscope may have interfered with the subject's performance, and single determinations of heart rate would have ignored much of the trial data while increasing the variability of measurement.
Biofeedback Heart Rate Training during Exercise
111
Subjects were admonished from talking during the trials. They could breathe in any manner they wished, tense or relax muscles, or think any thought they pleased, but they were instructed not to close their eyes or to move in such a way that the EKG signal was interfered with. Postures obviously taken to minimize the amount of exercise, such as leaning over the rails of the treadmill and dragging the feet, were neither allowed nor observed. During the trials, the observer sat behind and to the right of the subject out of the subject's view (Fig. 1) and remained silent. Between trials, if asked, the observer informed the subjects about techniques other subjects had used and about possible clinical applications of the results of the study. Feedback Subjects were assigned randomly to either Group E - C (experimental -control) or Group C - E (control-experimental). Subjects in Group E - C received heart rate feedback during the first five sessions of the study but not during the second five sessions, i.e., they began the experiment in an experimental, feedback condition and ended in a control, no-feedback condition. Subjects in Group C - E received no feedback during the first five sessions but did receive feedback during the second five sessions, i.e., they began the experiment in a control condition and ended in an experimental condition. The eight subjects assigned to Group E - C were instructed to try to lower their heart rates during the trials. Of the 10 subjects assigned to Group C - E , 5 (Subgroup T) were instructed for the first five sessions to try to lower their heart rates and 5 (Subgroup N) were instructed to walk normally. For the second five sessions, all the subjects in Group C - E were instructed to try to lower their heart rates. The average ages of the subjects in the various groups were comparable (27 years old for Group E - C ; 26 for Group C - E ; 26 for Subgroup T; and 25 for Subgroup N), as was the sexual composition of the groups (5 males and 3 females for Group E - C ; 5 males and 5 females for Group C - E ; 3 males and 2 females for Subgroup T; and 2 males and 3 females for Subgroup N). Heart rate feedback was provided in four forms for each subject in an experimental condition, as shown schematically in Fig. 1. First, an amber light was illuminated at the end of each R - R interval below a preset criterion. The amber light remained illuminated through the ensuing R - R interval. No light signal followed R - R intervals which were above criterion. Illumination of the signal light, therefore, indicated success at control of heart rate. Second, an electronic display of beat-to-beat heart rate provided additional information beyond the binary feedback of the signal light,
112
Goldstein et al.
because it indicated the degree of deviation from the criterion rate. Third, the polygraph's cardiotachometer channel was pointed out to the subjects and positioned to be clearly visible to them (Fig. 1); this provided information about trends in heart rate over time. Fourth, the observer commented on the subject's performance after--but not duringmeach trial. The subjects chose which feedback form to attend to and could switch from one to another at any time. The general rationale for setting the criterion heart rate was that of shaping, i.e., reinforcing slight changes in performance in the appropriate direction at first and then gradually increasing the difficulty of the criterion as performance improved. The observer set the criterion manually after each trial so that the signal light lit for approximately 80°70 of the beats during the next trial. For the first trial of the day, criteria were adjusted at 1 minute into the trial in order to achieve the 80070 goal. In practice, setting the criterion heart rate at 1-2 bpm above the mean heart rate on the previous trial usually satisfied the 80°70 goal, as did setting the criterion heart rate for Trial 1 of each session at the mean heart rate for Trial 1 of the previous session. Subjects were informed each time a criterion change was made. No subject received feedback during the first trial of the first experimental session in order to obtain preexperimental baseline information on heart rate during exercise for all groups prior to feedback. The feedback provided in these various forms constituted the only reinforcement for performance in the study.
Variables and Data Analysis Independent variables in this experiment were: (1) feedback condition; (2) instructions (i.e., try to lower heart rate vs. to walk normally); (3) sessions; and (4) trials within sessions. Dependent variables included: (I) mean trial heart rate; (2) systolic blood pressure; (3) rate-pressure product, expressed as mmHg X bpm X 10~; (4) baseline mean heart rate; and (5) percent change in heart rate from the baseline to the rate for a given trial during the session. The experimental data were analyzed in two steps. First, analyses of variance (ANOVAs) were performed and then independent means t-tests comparing the experimental groups. The reason for this apprach was that the data analysis involved multiple between-group comparisons, and the probability that at least one of these would reveal a statistically significant difference by chance alone was unacceptably high. ANOVAs provided an assessment of the significance of the overall effects of feedback, sessions, etc., on the dependent variables, and obtaining significant effects in the ANOVAs permitted application of multiple t-tests to the data without fear of wrongly inferring significant differences between groups.
Biofeedback Heart Rate Training during Exercise
113
The type of ANOVA used was a repeated measures design, betweenwithin analysis of variance, with measures taken in the same subjects repeatedly (e.g., over sessions and trials), some independent variables applied within subjects (e.g., sessions and trials), and some others applied between subjects (e.g., feedback). The ANOVAs also permitted assessment of the significance of interaction effects in the data. For instance, if the "Feedback X Trials" interaction effect was significant, this would imply that the effect of trials on the dependent variable was significantly influenced by the feedback provision to the group under consideration. Although ANOVAs allow a number of general inferences from data, they do not pinpoint where the significant effects are. Independent means ,t-tests were, therefore, conducted to compare the experimental groups for each trial and session once significant ANOVAs had been obtained. Onetailed t-tests were used in calculating significance levels, because we wished to test a specifically unidirectional hypothesis, namely, that heart rate biofeedback produces decreases in mean heart rate, in systolic pressure, and in the rate-pressure product, an index of myocardial oxygen consumption. Throughout the study, p values of less than 0.050 defined statistical significance. RESULTS
Sessions 1 through 5 Baseline Heart Rate. Resting (i.e., no exercise) baseline mean heart rates did not differ significantly between Groups E - C and C - E for any of Sessions 1 through 5 (Table I). The resting baseline mean heart rate for Table L Baseline Heart Rate Data for Groups E - C and C - E Group E - C
*
Group C - E
Session
Mean HR
SD a
Mean HR
SD a
df
t
p
1 2 3 4 5
90.0 90.5 89.1 88.5 89,8
15.7 8.0 12.0 9.2 14.9
90.8 84.7 89.9 86.7 85.7
13.9 11.1 10.8 12.3 10.7
16 16 16 16 16
0.11 1.20 0.14 0.34 0.67
n.s. n.s. n.s. n.s. n.s.
6 7 8 9 10
94,4 83,6 84,5 83.9 84.5
13.2 10.0 13.9 14.6 11.4
95.2 92.5 93.1 89.8 93.5
8.2 8.8 11o9 7.1 10.6
16 16 16 16 16
0A6 2.00 1,46 1.13 1.73
n.s.
